This reader needs JavaScript for search, threads and tags. Recent tweets from @michaeljburry:
- 2026-10-08 — The market breadth indicator I follow gives a strong signal every few decades, and where we stand today:
https://michaeljburry.substack.com/p/short-thoughts-the-ai-cloud-oligopoly?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-10-08 — In this article I greatly expand on this tweet.
The market's grieving process has been interrupted the last couple decades by heavy doses of monetary and fiscal cocaine. So much so that it is now hard to tell if grief can ever find its position again.
Denial is perhaps most easily found in market breadth indicators, and I follow one in particular that seems to give a very strong signal every few decades.
<blockquote class="twitter-tweet"><p lang="en" dir="ltr">The stock market is quite obviously in its first stage of grief, denial. Per 2000 and 2008, this stage lasts 6-9 months.</p>— Cassandra Unchained (@michaeljburry) <a href="https://x.com/michaeljburry/status/2107345405797122459?ref_src=twsrc%5Etfw">October 6, 2026</a></blockquote> <script async src="https://platform.x.com/widgets.js" charset="utf-8"></script> ↗
- 2026-10-08 — I posted an X article expanding on the tweet.
https://x.com/michaeljburry/status/2108277211610456409?s=20 ↗
- 2026-10-08 — Short Thoughts: The AI Cloud Oligopoly, Stages of Grief & a Flashing Bear Market Indicator
Other Peoples’ Money, Market Breadth & Market Structure, The Difference This Time
Michael Burry Join the Chat See Trading Posts Read Articles
I have recently considered that the tech giants, but for Apple, are spending as if their lives depended on it when their lives clearly do not.
Monopolies almost by definition need not spend wildly as if their lives depended on it.
This deserved no small further contemplation, although rather quickly it dawned on me that they expect this just to be another monopoly rent they collect, that they collectively have the expectation the government will tolerate their oligopoly on AI.
Nay, will tolerate and fear it.
So the evil gambit is to create and to allow this AI to grow on their machines, where they will threaten to control both economic and political forces greater than humanity has ever seen, overwhelming any politician’s, or politicians’, might.
Why are OpenAI and Anthropic and Oracle also spending so much? Surely they are not monopolies?!?
No, but they intend to be part of the oligopoly. None can imagine breaking the dominance of a Microsoft, Amazon, Google, or Meta within each silo.
However, there is clearly an opportunity to join in that dominance under cover of a new paradigm. The power and riches expected by these companies are evidently so great that even Larry Ellison was seduced to bet his company on it.
This brings new meaning to technocracy and whether such is what we face. What is really scary is that other than the grab for market power, all this spending does not have a defined goal or economic end-point. As such, the consequences are likely unknowable to all. Yet they do it anyway.
The governments most interested in this technology seem to be the autocratic ones and the United States. A good set-up for a lot of drama, if not terror.
Thankfully, I give these companies less credit than all that. I grant them the craven desire to expand a monopoly, to join an oligopoly. But this is a want, not a need, and human thought is far too redundant for what is being built.
So this may be how they think, but that does not mean this is how it will go. Ultimately, compression will do its deed, and much of what is being spent will be just so much sunk cost.
Data Center Financing at Risk
Meanwhile, I am hearing of creaks and groans – and some loud snapping and crackling. I speak of valuation marks in the private equity and private credit arenas. We’ve moved shadow banking offshore and reinvented the role of insurance companies.
These noises are so loud that I can hear them way out here.
I therefore know that Wall Street is hearing them. Wall Street knows how valuable the PE/PC/Insurer financial pyramids are to the financing of the AI buildout.
Again, I have spent parts IV and V of the Heretic’s Guide, as well as much of a Short Thoughts and a Hystory Rhymes, on this idea of growing leverage in the buildout of AI data centers, which fundamentally conflicts with financing timelines.
Below I’ll let slip a slide from Part VI, still in development, which outlines how those creaks and groans will transmit back to the data center buildout. People say this time is different. This is how it is different. Not a good thing.
Not to mention, long-term rates are rising before the timelines have a chance to play out at all.
If Wall Street understands, and hears the same nasty noises I do from beneath the preternatural calm of equity markets, then why is the bull run to new highs so well-affirmed, well-accepted, and well-projected?
Please subscribe to Cassandra Unchained for the rest of the article and many more features backed by facts and data, Short Thoughts, Trading Posts, the best Chat on Substack, and Portpourri, a new portfolio review feature coming soon. Thank you for reading! ↗
- 2026-10-08 — A fun game in times like these is to go to the S&P 500 Index and see how many profitable companies one can buy with the bubble private company valuation.
The number that Anthropic’s valuation buys is 78, including Domino’s, Clorox, Smucker, Stanley Black & Decker, Deckers, lululemon, McCormick, Tractor Supply, NVR, Albemarle, FedEx Freight, News Corp, Alllian Energy, Kimco Realty, Hormel Foods, Clorox, Weyerhauser, DaVita, Zimmer Biomet, Lennox, Masco, A.O. Smith, MGM, Wynn, Brown-Forman, Norwegian Cruise Lines, Hungtington Ingalls, and many more.
The largest inflation adjusted valuation at pre-IPO of 1990-2000 was UPS at $119 billion, 92 years old at the time. That was 26 times earnings at a net margin of about 8.6%, 2.4x sales. ↗
- 2026-10-06 — The only reason the CEOs of these tech monopolies are spending wildly on data centers like they are is because they assume they will be granted an oligopoly that is too big too fail. ↗
- 2026-10-06 — The stock market is quite obviously in its first stage of grief, denial.
Per 2000 and 2008, this stage lasts 6-9 months. ↗
- 2026-10-06 — You will learn something. ↗
- 2026-10-05 — Death n' Roll RIP LG
Bringer of Light (2022 Remaster) https://youtu.be/hkA4oZ7LIJQ?si=fuYmQy3vspA6OQqS via @YouTube ↗
- 2026-10-05 — In the last couple of weeks I've published four pieces - a Hystory Rhymes, a Short Thoughts, and two installments of the Heretic's Guide.
All together,my argument is simply the challenge to undertand, and by understanding, to consider one's luck in some what has happened here, and how it might turn. ↗
- 2026-10-05 — Did you know Cassandra Unchained has a web site?
https://michaeljburry.substack.com/ ↗
- 2026-10-05 — Fraud is inevitable in the dark. I bring light. ↗
- 2026-10-05 — Real analysis that critics don't read. ↗
- 2026-10-05 — Obituary, the AC/DC of Death Metal.
@obituarytheband 4 decades in
OBITUARY - Dying of Everything (Official Music Video) https://youtu.be/H-4r3ZjRhts?si=acrl5zjL8hi7GrJs via @YouTube ↗
- 2026-10-04 — Too academic is code for ? ↗
- 2026-10-04 — https://michaeljburry.substack.com/p/dont-believe-your-lyin-eyes-gpu-depreciation?r=4repfn&utm_medium=ios
Think twice. ↗
- 2026-10-03 — If you missed it, 9200 words of analysis on the fan favorite stocks. I tried not to waste any. $MSFT $GOOG $ORCL $META $AMZN
https://michaeljburry.substack.com/p/capital-cycle-iq-and-the-forensic?r=4repfn&utm_medium=ios ↗
- 2026-10-02 — One more time for our September 2026 Charity of the Month ↗
- 2026-10-02 — Short selling is perniciously perilous.
Market timing is garishly garrulous.
Ignorance is bliss.
Still, one can learn something new every day.
This is new. ↗
- 2026-10-01 — Don’t Believe Your Lyin’ Eyes, GPU Depreciation & Useful Lives
History Rhymes: Nvidia & The Great Winfield, “We have all been here before.”
Welcome to the second in the History Rhymes series, where I use the past to illustrate the present and maybe even illuminate the future.
For reasons unknown, the most valuable company in the world is on the road promoting itself. On September 27th, as part of a larger investor presentation, Nvidia threw up a slide designed to answer critics who still might suggest Nvidia’s chips are depreciating quickly. Critics such as yours truly, and others.
The slide, “Nvidia AI Infrastructure Retains Value Beyond Accelerated Depreciation Schedules,” plots the purported retained value of its A100, H100 and B200 chips against a 5-year accelerated depreciation curve.
The implication being values of these Nvidia GPUs across generations are high enough to make the case that companies are OVER-depreciating their chip investments.
Fans of Nvidia responded with joy at this master stroke, finishing off the bears once and for all. One such celebrant, Tae Kim, literally wrote an excellent book on Nvidia, The Nvidia Way: Jensen Huang and the Making of a Tech Giant.
All of this flex almost seems geared toward one critic’s argument which launched a nagging line of inquiry over how fast such chips and their machinery should be written off.
I know this critic very well, humph humph.
He arrived in 1968 amidst applause from Warren Buffett, who heartily recommended the critic’s new book, “The Money Game” in his mid-year partnership letter.
“For a magnificent account of the current financial scene, you should hurry out and get a copy of ‘The Money Game’ by Adam Smith. It is loaded with insights and supreme wit.”
Perhaps nothing this critic Mr. Smith wrote about left a mark on market history like the day he arrived in the Great Winfield’s office earlier that year or late the prior.
The Great Winfield wasted no time and mockingly disparaged the critic for his antiquated views on investing.
“Look at him, framing questions about depreciation, about how fast these computers are written off…” declared the Great Winfield. “You can’t make any money with questions like that. They show you’re middle-aged.”
The Great Winfield, you see, is an old friend of the critic’s and not just that but a “tape-reading super-speculator.” He recognizes the best players are not the middle-aged, but rather those “who have not passed their twenty-ninth birthdays.”
“My solution to the current market,” the Great Winfield said, “Kids. This is a kids’ market. This is Billy the Kid, Johnny the Kid, and Sheldon the Kid.”
The three Kids stood up, without taking their eyes from the moving tape, shook hands, and called me “Sir” with respect. “Aren’t they cute?” the Great Winfield asked, “Aren’t they fuzzy? Look at them, like teddy bears. It’s their market. I have taken them on for the duration.”
“I give them a little stake, they find the stocks, and we split the profits,” he said. “Billy the Kid here started with five thousand dollars and has run it up over half a million in the last six months.”
“Wow!” the critic said, and he asked Billy the Kid how he did it.
“Computer leasing stocks, sir!” he said, like a cadet being quizzed by an upperclassman. “I buy the convertibles, bank them, and buy some more.”
The critic suggested perhaps a lot of borrowing was involved.
“Not too heavily, sir!” said Billy the Kid. “I put up at least three percent cash. When I am conservative, I put up five percent cash...”
Billy the Kid said he was in Leasco Data Processing, and Data Processing and Financial General, and Randolph Computer, and a couple of others I can’t remember, except that they all have “Data Processing” or “Computer” in the title…
“The need for computers is practically infinite,” said the Kid. “Leasing has proved the only way to sell them, and computer companies themselves do not have the capital. Therefore, earnings will be up a hundred percent this year, will double next year, and will double again the year after. The surface has barely been scratched. The rise has scarcely begun.”
The Great Winfield eyed the critic and said, “I know what he’s going to ask. He’s going to ask what makes a finance company worth fifty times earnings.”
Billy the Kid smiled tolerantly, well aware that the older generation has trouble figuring out the New Math, the New Economics, and the New Market.
The Great Winfield continued.
“This one will really take you back. Sheldon’s Western Oil Shale has gone from three to thirty.”
“Sir!” said Sheldon the Kid. “The Western United States is sitting on a pool of oil five times as big as all the known reserves in the world- shale oil. Technology is coming along fast. When it comes, Equity Oil can earn seven hundred and fifty dollars a share. It’s selling for twenty-four dollars.”
“The first commercial underground nuclear test is coming up. The possibilities are so big no one can comprehend them.”
“Shale oil! Shale oil! said the Great Winfield, “Takes you way back doesn’t it? I bet you can barely remember it.”
The Great Winfield exchange above, plucked from the pages of Adam Smith’s The Money Game, published nearly 60 years ago, proves human nature is more constant across the human experience than just about any other feature.
Below, Billy the Kid’s 1968 pitch against the 2026 tape. Squint away.
Please subscribe to Cassandra Unchained to access the rest of the article, the meat of the article. ↗
- 2026-09-30 — History Rhymes: Large Language Models Off to a Bad Start?
This 145-year-old case study – presented at the Smithsonian Institute no less - provides a potentially devastating critique of today’s Large Language Models and the spending behind them.
Michael Burry
SUBSCRIBE TO CASSANDRA UNCHAINED
Welcome to a new series on my Cassandra Unchained Substack, History Rhymes, where I bring key perspectives from the distant past to bear on present events.
Cassandra Unchained is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
While mining old newspapers on a quiet Saturday – a hobby of mine - I came upon a story from June 19, 1880, that I found relevant to our modern anxieties about AI.
It is the story of Melville Ballard, who, as a child without language, spied with his eyes a tree stump and asked himself if the first man rose out of it.
New York Times, Saturday, June 19, 1880
This 145-year-old case study – presented at the Smithsonian Institute no less - provides a potentially devastating critique of today’s Large Language Models and the spending behind them. With a simple human story, it boldly announced that complex thought exists in the silence before words.
Today, well into the 21st century, by putting language before the capacity for reason, we are not building intelligence; we are building an increasingly sophisticated mirror.
SUBSCRIBE TO CASSANDRA UNCHAINED
There are actually two stories of interest in that old newspaper. Let’s start with the one in the middle. This is Page 3 of this edition of the New York Times, and I see a story called Thought without Language.
Of course, Large Language Models, Small Language Models, and Reasoning are the topic du jour.
This the full story title is, “Thought without Language, The Narrative of a Deaf-Mute, His First Thoughts and Experiences.” As well, the story was originally published in the Washington Star on June 12, 1880.
The story concerns one Professor Samuel Porter, of the National Deaf-Mute College at Kendall Green, who presented a paper at the Smithsonian Institution. The paper title, “Is There Thought Without Language? Case of a Deaf Mute.”
SUBSCRIBE TO CASSANDRA UNCHAINED
At first discussion of deaf-mutes and children having no form of mental action that distinguishes them from brutes, well, understanding has changed a lot, and I was ready to dismiss.
The case study is of a teacher at the Columbia Institute for the Instruction of the Deaf and Dumb. This particular teacher, Melville Ballard, is also a deaf mute and a graduate of the National Deaf Mute College.
Mr. Ballard says that in his infancy he communicated with his parents and brothers by natural signs or pantomime. His father, believing that observation would help to develop his faculties, frequently took him riding.He continues that it was during a ride two or three years before he was initiated into the rudiments of written language that he began to ask himself the question, “How came the world into being?” and his curiosity was awakened as to what was the origin of human life, its first appearance, the cause of the existence of earth, sun, moon, and stars. At one time, seeing a large stump, he asked himself the question, “Is it possible that the first man that ever came into the world rose out of that stump? But that stump is only a remnant of a once magnificent tree; and how came that tree? Why, it came only by beginning to grow out of the ground, just like these little trees now coming up;” and he dismissed from his mind as absurd the connection between the origin of man and a decaying old stump.He had no idea of what suggested to him the question as to the origin of things, but he had gained ideas of the descent from parent to child, of the propagation of animals and the production of plants from seeds.The question that occurred to his mind was whence came the first man, the first animal, and the first plant at the remotest distance of time, when there was no man, no animal, no plant, since he knew all had a beginning and an end. He thought most upon man and the earth, and believed that man would be annihilated, and there was no resurrection beyond the grave.It was when he was 5 years old that he began to understand the descent from parent to child, and he was about 8 or 9 years old when he began to ask as to the origin of the universe. Of the form of the earth, he inferred from a look at a map of the two hemispheres that there were two immense disks of matter lying near each other; that the sun and moon were two round, flat plates of illuminating matter, for which he entertained a sort of reverence, and thought from their coming up and going down that there must be a certain something having power to govern their course.He believed the sun went into a hole at the west and came out at another at the east, traveling through a great tube in the earth, descending the same curve as it [the sun] seemed to describe in the sky. The stars seemed to be tiny lights studded in the sky. He relates how vainly he endeavored to comprehend this subject until he came to school, at 11 years of age.Before this he had been told by his mother that there was a mysterious being up in the sky, but when she could not answer his questions he gave it up in despair, feeling sorrowful that he could not obtain a definite idea of the mysterious living one up in the sky.For a year after his admission into the school for deaf-mutes, he learned a few sentences every Sunday, and though he studied these simple words, he never acquired any idea of them. He attended the chapel services, but they were almost unintelligible, owing to his imperfect knowledge of the sign language as employed in the institution. The second year he had a small catechism, containing a series of questions and answers.
The development of understanding through the application of language to the capacity for reason continues.
He was now able to understand the sign language employed by his instructors. It may be said that his inquiring disposition ought to have been satisfied. It was not so; for when he had learned of the creation of the universe by the one great ruling Spirit, he began to ask himself, Whence came the Creator? and set himself to inquiring after His nature and origin. While he revolved this question he asked himself, “Shall we ever know the nature of God and comprehend His infinity after we enter His kingdom?” and would it not be better for us to say, with the patriarch of old, “Canst thou by searching find out God?”
Professor Porter then lets the hammer fall, as it were, on that 1880 Smithsonian Institute audience.
Prof. Porter discusses this case, and notes that brutes may understand some words, tell some objects from others, etc. He says: “Allowing to the brutes the utmost that can be claimed for them, is it not still plain that man has faculties which we cannot conceive as developed out of, or as simply exaltations in degree of, anything that he possesses in common with the lower animals.* * * Whatever may be the similarity in the way in which the impressions are produced or in the structure of the organs, and whatever may be the dependence upon organic action—that is to say, however they may be allied physiologically—yet as sensations or perceptions, those of the eye are different in themselves and imply a special gift or power not implied in those of the ear, or the head, or the tongue. It is not thus with the acts of the reason as compared with the working of the lower faculties.* * * That the two have some elements in common does not prove them to be throughout of the same order or render it possible for one to be developed out of the other. And if the eye of the soul, the higher reason by which we look through the universe of things, cannot look in upon itself and clearly discern its own nature and processes, we ought not, therefore, forgetting what it does, to deny its essential superiority and to assimilate it to those lower and subsidiary faculties which we can bring under its scrutiny. That by which we understand all things must be essentially superior to anything else that is understood by it.
SUBSCRIBE TO CASSANDRA UNCHAINED
One of the presentation’s attendees notes, significantly, how Ballard’s eyes conveyed meaning perfectly, without misunderstanding, above all else.
One of the most interesting features of this meeting was Mr. Ballard, by signs, explaining how his mother informed him that he was going a long way to school, where he would read from a book, write and fold a letter, and send it to her, &c., and also, by pantomime, reciting how a hunter, after killing a squirrel, accidentally shot and killed himself. Mr. Ballard’s signs and gestures, with the expression of the eyes and face, conveyed his meaning perfectly to the audience, and, in the words of a member, the expression of the eye was language which could not be misunderstood.
Let us consider these two statements:
“That by which we understand all things must be essentially superior to anything else that is understood by it.”
“…in the words of a member, the expression of the eye was language which could not be misunderstood.”
Ergo,
Language without the Capacity for Reason fails at Understanding
Only with Capacity for Reason does Language unlock Understanding.
Understanding, fully realized, transcends Language.
By putting language first, LLMs build a primitive form of reason purely through logical inference, but this form of reason has been shown flawed and prone to hallucination due to limitations at the many ragged edges of knowledge.
The capacity for reason never existed. Therefore, language cannot scale through reason to understanding.
The professor suggests, in his work with deaf and mute people, he has discovered that a capacity for true reason must exist first, before language, so language can unlock understanding — the product of that capacity for true reason and language.
“The expression of the eye is the language which cannot be misunderstood.”
To wit, expression of the eye is what flawless understanding looks like, without the need for language.
Large Language Models, by putting language first, before the capacity for true reason, can never attain understanding.
If understanding truly transcends language, as this 144-year old presentation at the Smithsonian, suggested, we should find evidence of it rather easily today.
SUBSCRIBE TO CASSANDRA UNCHAINED
I can personally relate this to the study and practice of medicine. All through undergraduate premed classes and through much of medical school, deductive logic is the means by which the student organizes the vast field of medicine. Through the clinical years the art of medicine, the physical manifestation, the emotion, the human expertise is developed. And then, at some point during residency or early in the physician’s career, with volume involving all of the above, understanding is reached. All the parts connect together in a vast, complex web that allows expert complete patient care.
Two surgeons working a difficult head and neck cancer or trauma, or the nurses that work with the doctors, can at times communicate purely with their eyes, and full understanding is transmitted, actions taken because all parties involved have attained understanding, transcending logical inference and other primitive forms of reason, such as the memorization and puzzle-building of the early education of a health care professional.
The eye then provides the intuitive grasp of reality based on mutual understanding, which proceeded from the capacity for reason in the presence of language.
The Large Language Model, as well the Small Language Model, lives permanently in the middle. It can simulate reason, but without true capacity for reason, without eyes, without understanding.
Ballard’s Test: An entity does not possess capacity for understanding until reason is demonstrated in the absence of language.
This is a known flaw, a bad start. The original approach to AI was to generate a true capacity for reason first, but it was never realized, and the field pivoted to language first because it was easier.
This ‘bad start’ has led to a “parameter trap,” where brute-force language processing powered by zillions of power-hungry chips has become an incredibly ironic bottleneck.
SUBSCRIBE TO CASSANDRA UNCHAINED
As my conversation with Klarna’s Sebastian Siemiatkowski highlighted, the future lies in compression—leveraging ‘System 2’ reasoning-first to work off the redundancy of information and the relatively finite query sets produced by humans to drastically reduce compute needs.
This new line rejects singularity through language models talking to each other in an infinite mirror as a directionless waste of resources made impossible by lack of a basis in economic realities.
While frontiers like Google’s AlphaGeometry and Meta’s Coconut are finally moving toward this ‘reason-first’ architecture, they are essentially rediscovering what was presented at the Smithsonian 144 years ago: that language is the output of understanding, not the engine of reason.
This multi-trillion dollar scaling myth may be rendered redundant by a return to the silence of pre-linguistic reason, the full-bandwidth capacity for reasoning of the deaf mute, whose silent thoughts reached for heaven’s stars before ever finding the words for them.
Silly Valley
Now, I mentioned there was another story of interest, and it is on the same page. More relevant to the first story than anyone in 1880s may have guessed it would be in 2026.
This article is “San Francisco’s Wealth, A Population of Bonanza Speculators.”
This story was written June 1 in San Francisco, and only published in the New York Times on June 19th.
SUBSCRIBE TO CASSANDRA UNCHAINED
The French saying, “the more things change, the more they stay the same,” comes to mind.
“Hard times” in San Francisco is a phrase that would mean almost “flush times” in an Eastern City, and means the absence of free-handed living and extravagant luxury rather than the presence of pinching poverty and cramped means.”California was pre-eminently the paradise of the man of small capital. To satisfy the craving for speculation, the peculiar open-board system was adopted, whereby the man who had $50 to invest, by purchasing a share therein, could acquire a small interest in a mine at a dollar a share, or two shares at 50 cents, or any number at varying prices.A “boom” existed here in certain stocks, seemed not to reach beyond the desire to do so “just once more” it seemed to excite the same gambling fever in San Francisco, and for lines lost by the bonanza firm was eagerly grasped by the people of San Francisco, and of the “boom” having been accompanied and by speculative losses on the part of the people, the “boom” disappeared and stocks fell to their normal condition.
The story closing hits hard for reality today.
The People of San Francisco seem to have become educated to the idea that they must leap into fortune at once, and their big bonanza at Virginia City having failed, they do appear to be willing to exert themselves to hunt for wealth in other directions, such as the development of manufacturing, trade, and agricultural interests. Almost the entire population is imbued with the passion for speculation, and if a new bonanza as big as the one in Nevada were to be discovered either there or near here, stocks would mount again to absurd figures, and San Francisco would again pass through the period of flush times to again suffer as she has during the past two years.
In The Cardinal Sign of a Bubble, Supply Side Gluttony, I outlined this remarkable tendency, originating in the San Francisco Bay Area, for speculation rising to levels that spur investment far, far beyond that needed for end demand in any timely dimension.
Reading old newspapers like this provides the perspective to interpret today’s events in a differentiated manner. Whether Silicon Valley will “again pass through the period of flush times to again suffer,” as she has repeatedly, or improbably break the pattern, I hope you found this helpful.
SUBSCRIBE TO CASSANDRA UNCHAINED ↗
- 2026-09-30 — link: https://michaeljburry.substack.com/p/history-rhymes-large-language-models?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web
History Rhymes: Large Language Models Off to a Bad Start? ↗
- 2026-09-30 — My protege! @philcliftonX wrote a terrific letter to his Pomerium Capital investors.
@Bloomberg is smart to read and report.
https://www.bloomberg.com/news/articles/2026-09-30/michael-burry-s-protege-warns-of-value-traps-in-korean-stocks ↗
- 2026-09-29 — Go Grok! ↗
- 2026-09-29 — For the benefit of humanity, the markets should tank hard and prevent the OpenAI and Anthropic IPOs. ↗
- 2026-09-29 — Take your mind off the market for a bit. ↗
- 2026-09-28 — web page : https://michaeljburry.substack.com/ ↗
- 2026-09-28 — link: https://michaeljburry.substack.com/p/trading-post-september-28-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-09-28 — New Trading Post September 28, 2028
Big Action in Big Shorts As My Research Progresses, plus a few new Buys
My work on Part VI of the Heretic's Guide dovetails with Ares’s work wonderfully. I am coming at it from the bottoms-up insurance side, and I have collected significant data and come to certain conclusions. My data is more up to date than the data Ares uses. As well, my data is greater in scope on the insurance side. ↗
- 2026-09-28 — If you missed it, this post will advance understanding of what's going on in the world of AI by a ton. ↗
- 2026-09-28 — link https://michaeljburry.substack.com/p/short-thoughts-a-wall-street-titan?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-09-28 — Short Thoughts: A Wall Street Titan Hands Us Part V.5 of Our Heretic's Guide to AI
Researching for Heretic’s Guide Part VI, I Stumble Upon Can’t-Wait Info That is Must Know Now
Read → ↗
- 2026-09-27 — If not the weekend, now is the time. ↗
- 2026-09-27 — If you live in #California read this free PSA research and look into whether it is happening in your neighborhood. #calfire ↗
- 2026-09-27 — This is to my point about compression being inevitable as human knowledge is too small for what we are building. As well, humans are too redundant in their wants needs and questions.
AI-generated content will clearly contain propagation errors just like human history of knowledge does. Only LLMs will iterate those propagation errors infinitely faster with less ability to self- correct, for want of understanding.
This gets to Ballard’s test. LLMs cannot attain understanding (AGI) as understanding cannot exist unless reason first exists without language. A likely impossibility for a language model.
Research on this is already focused on getting around this in some way. Though many also have not yet conceded the point. ↗
- 2026-09-26 — The weekend is the perfect time to read a full 9200 word feature article.
Abridged - Capital Cycle IQ & Forensic Files on the Big 5 Hyperscalers (MSFT, AMZN, ORCL, META, GOOG)
However, for those with less time, or from around the world, that have requested the Abridged version, here is the 1750 word version of the feature, more directly stated in general.
With 4 out of every 5 words gone, the substance is present but the personality and the full bandwidth of the arguments are clearly lost. I encourage everyone to spend some time with the longer 9200 word version if possible. Thank you for reading! ↗
- 2026-09-25 — Note that Meta's Residual Value Guarantee is JUNIOR to this 2049 note ↗
- 2026-09-25 — 7.4% through 2049 if anyone wants a piece of data center debt kinda backed by Meta... ↗
- 2026-09-25 — Forensic Analysis on $GOOG, $META, $ORCL, $AMZN, and $MSFT
The Deepest Dives on the Big 5 Hyperscalers.
https://michaeljburry.substack.com/p/capital-cycle-iq-and-the-forensic?r=4repfn&utm_medium=ios&shareImageVariant=solid
Must know tactical information for investors in the indices they dominate as well as the stocks themselves ↗
- 2026-09-25 — New post, a long one. Deep forensic dive on $MSFT, $GOOG, $ORCL, $META, $AMZN with CU Universe stock rankings and the return of the fat pitches AICT Composite Value Map.
https://michaeljburry.substack.com/p/capital-cycle-iq-and-the-forensic?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-09-24 — Capital Cycle IQ & the Forensic Files of the Big 5 Hyperscalers
The Heretic’s Guide to AI’s Stars Part V, with CU Universe Rankings, the All Map
https://michaeljburry.substack.com/
The Heretic’s Guide to AI’s Stars Part V, with CU Universe Rankings, the All Map
Where are we in the capital cycle? Every investor should know about this, even if it is not folded into one’s analysis of stocks or other investments.
The capital cycle explains a tremendous amount about the biggest moves in asset prices over the last 3 decades.
Value investing has been abandoned almost wholesale by the market, they say. This is not your father’s market, they say.
But it really is. We are just smack dab in the middle of the capital cycle.
If you have been around since the beginning of Cassandra Unchained, you may know what being in the middle of the capital cycle actually means.
That is a bit busy, but let’s go through it. Blow it up. The above chart charts the percentage of net capital investment to GDP in green bars. Those green bars happen to reveal a pattern that we call the capital cycle.
The white horizontal line up there is the current net investment (Capital Expenditures less Depreciation) of the S&P 500 companies divided by Nominal GDP. It comes to about 2.07%.
2.07% is higher net investment to GDP than at any time in any prior capital cycle over the last nearly four decades, but for the tech-media-telecom bubble aka dot com bubble of the late 1990s.
Trump should be all over this.
But actually, I have not been precise. Today’s level is higher than any prior time save for the aftermath of the massive NASDAQ market peak in March 2000, from which it did not fully recover for a decade and a half.
Ah, that’s why Trump is instead all over anyone calling the AI phenomenon a bubble.
Consider that aftermath, when capital expenditures kept growing well past the market peak. Looking at the area under an imaginary curve there, it seems the NASDAQ peaked almost smack dab in the middle of that massive capital cycle.
Later, it turns out the S&P 500 and all its financials and real estate businesses peaked in November 2007. But look at the green bars.
That market peak in late 2007, which was also a last hurrah for the NASDAQ for a good number more years, occurred well below the bulk of net capital investment in that housing-dominated capital cycle.
Believe it or not, investment continued to ramp right through 2009, as banks failed and markets fell apart.
We also see in violet the trace left by EQT, a major gas and shale company, as it peaked with the S&P Energy Index a couple quarters before investment peaked. Once again, the shale boom-led capital cycle of the 2010s saw energy stocks peak in the middle of that boom’s net investment curve.
As of June 30th, the last data available, we are at that 2.07%. This coincided with a new all-time high for the NASDAQ. Not much to celebrate, as that June all-time high was the 94th over the prior three years.
Another day at the office, though here we are in September and no all-time high in the NASDAQ since June.
I have little doubt the next few quarters will set still higher and higher net investment/GDP marks, possibly even eclipsing that aftermath of the 2000 tech stock peak.
Let’s look at the same chart, stripped down.
Let’s take a moment to note those 12 quarters from the middle of 2003 to the middle of 2006 where net investment was negative.
What does that mean? Well, it means depreciation absolutely overloaded and overwhelmed capital expenditures for those twelve straight quarters.
Why do I look at depreciation? Because of this. Hold that thought, as this will be relevant later here.
The hangover from the TMT buildout, in the form of absolutely massive depreciation expense and ongoing write-downs of capital equipment, was so significantly negative that it dragged the S&P 500’s aggregate net investment negative during the rise of a housing bubble in the US so very great that it would take down and/or see the disappearance of names such as Merrill Lynch, Lehman, Bear Stearns, Washington Mutual, Countrywide, AIG, Wachovia, New Century, IndyMac, and, almost, Goldman Sachs.
That is how bad the bust before the housing bust was.
So, are we again in the middle of a capital cycle? Yes.
Will I continue on this idea that depreciation expense is being understated and write-downs are the future? Yes.
Why? Because of the capital cycle.
But I need more info, always.
We have already started, of course.
In Part IV, we looked at some scary aggregate numbers emanating from the Big 5 public hyperscalers - Microsoft, Amazon, Alphabet, Meta, and Oracle.
That $3 trillion in aggregate purchase commitments, future leases and guarantees backing third-party debt, construction-in-progress and various SPVs - these all point to epic levels of spending over the next few years.
The most profitable companies in the world, except Apple, are betting everything on this as their free cash flow turns negative and the borrowing pace increases.
There are signs of strain at each of the big hyperscalers, and recent earnings reports, filings and conference calls provide clues as to how stressed each one really is.
To receive new articles, short thoughts, trading posts, and access to the best chat on Substack, consider becoming a free or paid subscriber.
The Hyperscalers, a Forensic Check-Up
When the write-offs come, perhaps in 2028 or 2029, these commitments discussed in Part IV may be so large that a relatively small write-off has a bigger impact than we can now imagine. After all, the growth rate of these off-balance sheet commitments and exposures is eye-watering.
But there is more to these companies. Let’s take a forensic look at what the filings, earnings releases and earnings calls revealed for each.
There is so much more! To read visit https://michaeljburry.substack.com/
A forensic deep dive through the Big 5's regulatory filings, with valuations, rankings of all stocks in the CU covered universe, and more.
https://michaeljburry.substack.com/ ↗
- 2026-09-23 — The State of California is Incubating the Next Big Deadly Fire Nightmare Right Now.
New Article Explains All Right Here on X.
It is time to fight for public safety.
https://saratogacred.org/get-involved#donate
The City’s Own Report Demonstrates People Could Die.
The City, Fearing the State and Developers, Approves Projects Anyway.
We Sued. The petition: http://saratogacred.org
#NeverForget #californiafires
This is, ironically, developing and building, parcel by parcel, the exact configuration and circumstances that killed 116 people and erased 35,000 homes and structures in Paradise, Pacific Palisades and Altadena.
This is playing out all over the state. The seeds are being planted for the next horrific deadly fire nighmare, and the next, and the next, and the next... ↗
- 2026-09-23 — The Next Big Deadly Fire Nightmare is Being Created Right Now by the State of California
The City's Own Report Demonstrates People Could Die.
The City, Fearing the State, Approved the Project Anyway.
We Sued. The petition: saratogacred.org
Never Forget
On November 8, 2018, the Camp Fire overran Paradise, California. At its peak, the fire consumed the equivalent of one football field every second. Eighty-five people died, some in their cars, trapped in the gridlock created by poor escape routes.
There were just too few roads out of town while the fire moved faster than traffic. In addition to the 85 people who horrifically lost their lives, nearly 19,000 homes and structures burned.
Have we already forgot? On January 7, 2025, the Palisades Fire ignited at about 10:30 a.m. on about 10 acres. Within twenty minutes, the fire grew to 200 acres. A hellscape ensued.
Fire outran evacuation almost immediately. Residents became stuck in gridlock on Sunset Boulevard and had to abandon their cars to flee on foot. Bulldozers were then needed to shove hundreds of abandoned vehicles off the road so the fire engines could get through.
That same day, the Eaton fire erupted in Altadena. It would kill 19 people and take down 9,416 structures.
Across the Palisades and Eaton fires, 31 people lost their lives, and more than 16,000 structures were destroyed. 100,000 people were displaced.
The one common thread for all three fires was overdevelopment in fire-prone hills. Hills that were designated very high fire hazard zones. In each case, narrow evacuation routes jammed, and the fires moved faster than 116 people could run, leading to unimaginable horrors.
35,000 homes and structures were burned down.
After the Paradise fire in 2018, the California Legislature passed AB 747, which required cities to analyze whether their roads could handle mass evacuation.
The bill said it plain. Planning for mass evacuation in advance "will help save lives." The lesson is literally written into state law.
That is, before adding people and structures to a fire-prone hillside, prove that they can get out faster than the fire consumes the area..
The city of Saratoga, California did the study. The study said the windy two lane road out of the mountains, Pierce Road, already fails as an evacuation route.
Residents will take five hours to get out, the study said
The fire will take four. That should make everyone’s heart skip a beat.
Nevertheless, August 19, the Council approved a 25-home subdivision called Masson Estates on an undeveloped, wooded hillside that CAL FIRE designates a Very High Fire Hazard Severity Zone.
Before voting yes, the Council adopted this finding,
"The proposed Project would cause irreversible environmental changes associated with its significant and unavoidable impacts, namely: impacts related to… emergency access, emergency response and evacuation, wildfire hazards, consistency with an emergency response/evacuation plan, and exposure of project occupants to pollutant concentrations from wildfire."
The City knew the danger, wrote it down, and approved the project anyway.
On September 18, our 501(c)(3) nonprofit, Citizens for Responsible and Equitable Development (CRED), filed suit in Santa Clara County Superior Court. CRED v. City of Saratoga, Case No. 26CV503751. The verified petition is at saratogacred.org.
The Record, Briefly
The site sits at Pierce Road and Saratoga Heights Drive. Pierce Road is a narrow, curving two-lane mountain road with steep grades, blind curves and minimal shoulders. Residents of the foothills must travel this road every day and are familiar with it.
I live at the base of Pierce Road. I see them come and go all day long, in orderly if busy fashion. I have been back there, up there to visit friends. I know what is up there.
A tinderbox.
Again, the City of Saratoga’s own modeling shows Pierce Road cannot handle current evacuation demand. Today, without any new projects. They know this, and approved new development anyway.
How marginal is this road?
Almost unbelievably bad as a road, let alone as an escape route.
The City's own Public Works Department, reviewing this very application in January 2024, required the developer to design and build a stabilizing structure for "existing ground movement.”
Yes, Public Works called out an active landslide on Pierce Road along the project's own frontage., requiring a storm drain redesign with full width reinforcement above.
Too, the City's own evacuation study (Fehr & Peers, May 2026) projected the project would add roughly 30 minutes to evacuation time in the event of a fire. This is a 13% increase over baseline.
KLD, a nationally recognized evacuation expert, reviewed that study at our behest and concluded the extra delay could prove pivotal to escaping residents.
The delay "directly increase[s] the likelihood of vehicles becoming trapped on constrained corridors like Pierce Road, creating a critical life-safety hazard and significantly elevating the potential for catastrophic loss of life."
Vehicles trapped on a constrained corridor do not exist only in these modeling exercises.
We have the evidence, the 85 lives lost in Paradise, the hundreds of cars that needed to be bulldozed off Sunset Boulevard.
Pierce Road cannot be any better and appears even worse, now, than those escape routes were before their fires.
I am a UCLA Bruin, and I can tell you firsthand, Pierce Road is no Sunset Boulevard.
KLD’s study concluded that allowing dense development along Pierce Road and the branches off it up into the hills, under the modeled fire conditions, "creates a significant threat to human life," and that these delays are "major, life-threatening impacts that must be fully addressed before any further development approvals are granted."
On July 8, the City's own Planning Commission denied the project as benefits were not found to outweigh the evacuation risk. The developer, of course, appealed.
Six weeks later, the City Council reversed their decision.
Council members stated on the record they approved the project due to fear of legal retribution from either the State’s housing agency or the developer or both.
Surprise, in America, fear of the State is not a legal reason for a City Council to approve a project unsafe for existing residents.
Sixteen days after approval, on September 4, a brush fire broke out on Mt. Eden Road, 1.4 miles from the site. Winds were light. It was contained. Winds are not always so light, especially during fire season.
The City Council’s fear of legal retribution is simply not the law of the land.
The Housing Accountability Act itself - Gov. Code § 65589.5(d)(2)- permits a city to deny a housing project that would have a specific, adverse, unavoidable impact on public health or safety.
Section 65589.5(e) expressly preserves California Environmental Quality Act (CEQA), and the predicate findings are the City’s owns: the certified Environmental Impact Report (EIR), and the commissioned Fehr & Peers evacuation study.
The City had the power to say no. It said yes. In doing so, it called non sequitur remediation such as a public trail and a retaining wall "overriding considerations" that outweighed documented risks to human life.
Our lawsuit recites what the fact pattern and record say in their entirety.
The approval violates CEQA, which California courts have repeatedly held requires honest analysis of wildfire evacuation before homes go into the wildland-urban interface.
It violates the Subdivision Map Act, which requires denial of a subdivision likely to cause serious public health problems (Gov. Code § 66474(f)).
An unnecessary extreme fire hazard in a residential community is a public nuisance.
The development poses a threat to life from fire according to the City itself and therefore violates Article I, Section 1 of the California Constitution, which names "obtaining safety" among the inalienable rights of all citizens. Even Saratoga mountain dwellers.
What "Affordable Housing" Means Here
First, honesty about the affordability crisis, because it is real. The Bay Area has priced out its own teachers, nurses, firefighters and service workers. These are people that often commute hours from the remote Central Valley because no housing near their jobs is within reach. Anyone who dismisses that is not serious. When my son worked at Applied Materials in Santa Clara, a fellow worker of his was living out of his car in the parking lot.
We founded CRED in part to insist that new housing here be truly affordable. And we judge these hillside projects by that standard.
Masson Estates is a builder's remedy project. This means that because Saratoga's housing element was not certified by the State of Calfornia in July 2023, developers from all over could claim the City lost the power to apply its own zoning rules. That is some valuable immunity that developers have long sought and never quite achieved.
Today, that immunity is in most circumstances state-sanctioned, as long as twenty percent of the homes are “affordable” units. This is the core problem, a fraught question, and it starts with the definition of “affordable.”
At Masson, five homes out of the 25 will be “affordable” and squeezed on lots as small as ~2100 square feet, while the 20 market-rate homes run to 7,000+ square feet on lots up to two acres.
Interestingly, the application grew from 21 homes to 25 between the preliminary and formal filings; SB 330 voids vested rights above a 20% increase, and 21 × 1.2 = 25.2. The developer is building to the regulatory maximum number of units.
All bars on that visual above are on the same scale.
Take a moment and draw that in.
Now look at what "affordable" means in those 5 homes. State law defines the tier as "lower-income households" which is up to 80% of area median income. The median family income in the San Jose/Sunnyvale/Santa Clara metro area is $205,500.
This is the highest in America because this area and the surrounding foothills is the Valley in Silicon Valley.
Saratoga along with Los Gatos and Cupertino are in the foothills of the Santa Cruz mountains and all are roughly half in half out of the mountains.
Saratoga has plenty of open space not in the mountains. Saratoga tried to build out equitable affordable housing there, but the State said no.
The outside developers that long wanted a piece of Silicon Valley wanted to develop luxury homes in the mountains most of all. The State of California saw an opportunity and arranged that, on behalf of the developers. Builder's Remedy.
In any event, "low income" here in Silicon Valley means a family of four earning up to $162,400, which is double the median American household's income. Low income is also a single earner making $113,700 or less, also well above the national average.
The capped housing cost for a qualifying low income family runs near 30% of that income which at the top end is about $4,000 a month.
The Masson Estates “affordable” housing will be ~2400 square foot four-bedroom houses on prime real estate with a view. These are not modest units, but they are "affordable housing" that will sell for $3-$4 million per.
The developers will make out wonderfully, with that kind of home going on a lot slightly smaller than the house at 2100 square feet.
And that is what this is. A boondoggle for developers of unaffordable housing. State sanctioned with full immunity from zoning laws in the name of equity. That is not equity, and it is time someone called them out on it. Building multi-million dollar homes in very high fire hazard zones at the expense of residentts' safety is no moral high ground..
Existing Saratoga homes, which are mostly 4-5 decades old, tend to sell at a median of $1,510 per square foot. At a planned 7,091 square feet, that pencils out to roughly $10.7 million, which is about 66 times the "low income" qualifying income, or 133 years of the median American household's gross pay.
The twenty market-rate homes together will likely hit near or above $100 million, and selling five slightly less expensive but still roughly $3.6 million dollar homes is the entire “price” paid by the developers and builders here.
For the very-low and extremely-low income tiers, these projects these deliver zero units. Not fewer. Zero.
If the goal were affordable housing, you would build it where affordable housing works: the valley floor, near transit, main roads, jobs, schools and fire stations. The “builder's remedy” law as practiced in Saratoga's hills is not an affordable housing program. It is a mansion-entitlement program sited where the state's own fire maps say no one should add people.
Masson is Not Alone on This Road
Documents CRED obtained through Public Records Act requests show details behind what's queued for development in the same mountains with the same narrow escape road.
To wit, the Masson Estates developer is co-owner of the Mountain Winery in the same area, but just outside Saratoga’s limits. Still in the very high fire risk area, the developer is looking at a 255-unit, 81-room-hotel redevelopment of the winery property. Construction has begun.
At 12991 Pierce Road, roughly two dozen three-story homes are proposed on a 1.96-acre parcel dropping to Calabazas Creek. The developer is claiming they should be entitled to a CEQA "infill" exemption and subject to no environmental review at all.
On Mt. Eden Road, where the September 4 brush fire broke out, a 7-unit builder's remedy application was deemed complete in December 2024.
Then there is Chadwick Heights, a Builder’s Remedy project of 97 units of dense housing - fourplexes, townhomes, duplexes - proposed on 25 hillside acres that the City's own geotechnical consultant described as essentially an ancient landslide inside a landslide hazard zone. There are actually active visible slides on the property.
The developer's solution is to move over 800,000 cubic yards of earth with cuts 90 feet deep and fills 70 feet high to prop up the landslide it would build on. This is also to divert creeks on site into culverts, as protecting the creeks is “not feasible given the amount of grading.” , and to run the site's creeks through culverts, because protecting them is "not feasible given the amount of grading."
Saratoga's municipal code limits hillside grading to 1,000 cubic yards.
800,000.
There are others.
It is a free for all thanks to fear of the State of California.
The City's own study says Pierce Road fails at today's population. Masson Estates' 70 new evacuees are the down payment on a corridor with more than 400 units queued behind them. A free for all.
Why We Fight
We established CRED to give residents a voice against overdevelopment in the fire-prone Saratoga hills. The affordability crisis cannot be solved by siting new tinderboxes in Very High Fire Hazard Severity Zones at the top of evacuation routes that already are failing and projected to fail in the event of a fire.
This is, ironically, developing and building, parcel by parcel, the exact configuration and circumstances that killed 116 people and erased 35,000 homes and structures in Paradise, Pacific Palisades and Altadena.
The City's own EIR also found significant, unavoidable danger to the future occupants of the project itself.
As well, the people most endangered in any evacuation are those without cars, dependent on transit that will not come up a gridlocked mountain road while the hillside burns.
City Hall fears developers. It fears the State. It fears well-funded pressure groups.
Fear of the State has displaced Saratoga's duty to its own residents, who are left holding their constitutional right to safety as the soft target.
This is playing out all over the state. I have no doubt, from what I am seeing here, that the seeds are being planted for the next horrific deadly fire nighmare, and the next, and the next, and the next...
For 18 months, across numerous City Council meetings, CRED made its case in public.
The lawsuit now speaks for itself.
The petition: saratogacred.org
Michael Burry, M.D., President, Citizens for Responsible and Equitable Development (CRED) ↗
- 2026-09-22 — “Regarding memory chips—which have seen the sharpest price surges this year—Chen observed that while many industry players continue to publicly predict price increases extending through the end of next year, the reality is that inventories are actually overflowing. Suppliers from mainland China are flooding the market with volume, acting as price disruptors; as for this period of price volatility, "those in the know understand the situation."”
Acer CEO says memory chip supply is rising. Remember the shortage in normal RAM was created by repurposing for HDM but now the production for normal RAM is ramping up - as it always does .
https://money.udn.com/money/story/5612/9764832 ↗
- 2026-09-21 — The crawfish are no match for this trick, and soon I have a handful.
I put them in my pocket.
Seeking.
I leave the tadpoles be.
I think of copper, and how it gets prettier as it ages.
Those tadpoles face a different fate, and they will eat flies.
https://michaeljburry.substack.com/p/trading-post-september-21-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-09-21 — The house party is packed, pushing AI higher today but I am largely ignoring the “woo-hoos.”
I’m out in the backyard with my flashlight, seeking.
I strike out at the salamander. Multicolored or just green, I blink and remain unsure. It blinks.
Perhaps from the ravine.
I fiddle in my pocket for the zinc nut and a bit of string. ↗
- 2026-09-21 — My Market Structure articles explain fairly clearly why the markets do what they do, and how they might change. ↗
- 2026-09-21 — Foundations: Market Structure, Volatility Targeting, Pod Shops & Other Gremlins
What this says is that the market is picking itself up more day-to-day than ever before. In fact, 40 years ago, the market just randomly walked through one day to the next. Now, there is a mean reversion, a walk back to the starting point.
In fact, within and especially across days, this market increasingly reverses itself.
The 1985-2000 market was 1.02, a coin flip, a random walk day to day.
Not anymore. 0.881, nope.
Again for the academics, the two-day variance ratio today is 0.881 with a z-statistic of −2.56 across 3,774 sessions. That’s walking back.
How the Market Works Today, Yesterday, and the Decades before THAT. ↗
- 2026-09-20 — The Heretic’s Guide to AI’s Stars Part IV: The Big 5 Hyperscalers & the Missing $3 Trillion
Current GPU pricing is representative of today’s scarcity, but the duration mismatch playing out undermines any guarantee the demand will continue at this level for long enough to justify today’s valuations.
I am a fair-weather fan of Nvidia’s names for its GPUs, but the newest line of chips I’ve dubbed LEAPFROG.
https://michaeljburry.substack.com/p/the-heretics-guide-to-ais-stars-part-c9c?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web
A study of overlooked footnotes to the 10-K and 10-Q filings from Amazon, Meta, Alphabet, Microsoft & Oracle ↗
- 2026-09-20 — The web site also gives a big picture overview of the types of content on Cassandra Unchained. ↗
- 2026-09-20 — Shout out to Precision Guitar Kits @GuitarKitsforU
This Honduran Mahogany explorer body blows me away. Rings like a big brass bell, just handling it you can hear it wants to roar like @orbitculture ↗
- 2026-09-19 — Well the 10 year went from 50bps to 500bps and stocks tripled. The 30 yr fixed mortgage went from 2,5% to 7% and home equity went from $20 trillion to more than
$30 trillion. Jeffrey sees it clearly. We are in backward land. ↗
- 2026-09-19 — Remember there were several books out about the 1990s bubble during 1999 including the Devil Take the Hindmost by Edward Chancellor, the New New Thing by Michael Lewis, and The Internet Bubble by the two Perkins’s. So yes a few of us are being loud about what is going on today ↗
- 2026-09-19 — Topical once more. ↗
- 2026-09-19 — Subscriptions to Cassandra Unchained through the web site avoid significant in-app fees.
https://michaeljburry.substack.com
Or, subscribe for free to be notified when I drop Articles, Trading Posts, Short Thoughts, Notes, and when I start new Chat discussions, like I am about to do right now. ↗
- 2026-09-19 — Here is the post. Had been waiting for Oracle and she did not disappoint.
https://michaeljburry.substack.com/p/the-heretics-guide-to-ais-stars-part-c9c?r=4repfn&utm_medium=ios ↗
- 2026-09-19 — New addition to my Heretic’s Guide series on Substack, dropped at midnight EST . A must read this weekend. Part V to drop in a few days, and I have Part VI lined up. ↗
- 2026-09-19 — Obituary. The AC/DC of Death Metal.
https://youtu.be/DSZINxB6Vvg?si=9WrWhEpb82pcCfK1 @obituarytheband ↗
- 2026-09-18 — Please donate to the Vanderbilt-Meharry Center for Excellence in Sickle Cell Disease Sickle Cell Development Fund https://give.vanderbilthealth.org/sicklecell
The Sickle Cell Development Fund has been selected as the Cassandra Unchained September 2026 Charity of the Month. ↗
- 2026-09-18 — Please donate to the Vanderbilt-Meharry Centr for Excellence in Sickle Cell Disease Sickle Cell Development Fund https://give.vanderbilthealth.org/sicklecell ↗
- 2026-09-17 — The Sickle Cell Development Fund - Cassandra Unchained September 2026 Charity of the Month
Please Donate Now to The Sickle Cell Development Fund
One of my starkest memories as an internal medicine intern is of patients literally lining the walls adjacent to the Elvis Presley Memorial Trauma Center in Memphis, Tennessee.
These patients were sitting down, dozens of them. They all were doing the same thing.
Breathing.
You see, these were sickle cell patients, and the oxygen that these patients were inhaling through masks was providing needed relief.
Oxygen was no cure, nor even very effective relief.
But some relief, any relief, clearly comforted these patients.
One does not see this in every hospital, but it brought home for me the disease that is sickle cell.
In sickle cell disease, patients’ red blood cells carry a single letter mutation in the hemoglobin gene. which produces what we call hemoglobin S.
When hemoglobin S releases oxygen, as it will in the low-oxygen environments of capillaries and small veins (the end of the road for oxygen delivery), it converts red blood cells into rigid, sharp crescent shapes. Like sickles. These then pile up like sticks and jam those small vessels.
Tissue downstream of the pile-up becomes starved of oxygen, causing more red blood cells to sickle.
The disease feeds itself. These attacks are crushingly painful, recurrent, and lifelong.
A cure now exists. Bone marrow transplantation has been curative for decades, but the cost is hundreds of thousands of dollars, often more with complications.
Beyond cost, the barriers include donor matching and all the risks associated with transplant. The process of bone marrow transplantation is probably the most stunning thing I saw in medicine. The transformation to a cure is remarkable, but the gauntlet to be run is challenging and not without risk.
New gene therapies were approved in 2023, but these are only available in a handful of Western health centers and only at great cost. List prices are from $2.2 million to $3.1 million per patient.
Over the past decade, the Vanderbilt-Meharry Center of Excellence in Sickle Cell Disease, led by Michael R. DeBaun, MD, MPH, has become a focal point for the treatment of sickle cell disease both inside the United States and abroad.
The majority of children born with sickle cell disease are born in sub-Saharan Africa, and in Nigeria especially, where roughly 150,000 are born with sickle cell annually. Bone marrow transplantation and gene therapy are not feasible at scale.
Therefore, Dr. DeBaun has led a heroic effort to prevent strokes in children with sickle cell disease in Nigeria. This involves screening with transcranial Doppler ultrasound and treating at-risk children with generic medicine to prevent sickling before the first stroke happens. The cost per child is dramatically less, and the dollar goes much further to helping these children.
Charitable financial support of this program will help Dr. DeBaun and his team reach many more children. Over 11% of children with sickle cell will have a stroke before age 20 absent intervention.
Please Donate Now to The Sickle Cell Development Fund
The gap between disease and cure is now an access problem. Dr. DeBaun has found a way to create access, and donations make that happen.
Donations to the Sickle Cell Development Fund also support research into delivering better care and cures for sickle cell.
Donations are tax deductible.
Memphis has one of the densest populations of sickle cell populations, and for a year I lived and treated patients there, I felt that need deeply.
Even now, I think of all the people around the world that have lived all this time with this disease, and it is heartbreaking.
Charitable giving and support helps transform bold research into sustainable breakthroughs for children and adults living with sickle cell disease in the U.S. and around the world.
Donate Now
Cassandra Unchained will donate 5% of its gross revenues for the month of September to the Vanderbilt-Meharry Center of Excellence in Sickle Cell Disease Sickle Cell Development Fund.
Thank you for your consideration.
Until Next Time! ↗
- 2026-09-17 — “One of the reasons why I am in favor of less government is because when you have more government, industrialists take it over, and the two together form a coalition against the ordinary worker and the ordinary consumer.”
— Milton Friedman ↗
- 2026-09-16 — I am righteously indignant over the wealth transfer to the very few that these bubbles create. The whole system is about creating bubbles so the grift can happen, both inside companies, transferring wealth to their employees at obscene rates, and in politics, as we see all over. ↗
- 2026-09-15 — Happy to announce that I have taken a role as Senior Advisor to Minerva Investment Management, led by Lakshmi Ganapathi and her global team at @UnicusResearch.
Unicus has built a formidable global reputation with hard-hitting credit analysis. I will enjoy my front row seat as Laks and her team break new ground with their short-biased investment fund. ↗
- 2026-09-15 — This is from my May article, The Heretic's Guide to Ai's Stars Part III.
I created the Pyramid to predict what would happen.
Today the Pyramid is showing what is happening.
https://michaeljburry.substack.com/p/the-heretics-guide-to-ais-stars-part?r=4repfn&utm_campaign=post-expanded-share&utm_medium=post%20viewer
The article also dug into financing for data centers, how it involves insurance companies and their captive offshore entitites. ↗
- 2026-09-15 — Bears on AI spending have asked, "how will all this be monetized?" The answer has been commercial enterprises are already doing that by buying compute, burning tokens, maxing out that compute, that token usage."
There is good reason for that, but perhaps it does not inform long-term demand.
So now bears ask "for how long?"
The pyramid below is not a static economic construct. There is strong pressure to migrate upward. Every enterprise that is profit-maximizing will seek to move up in the pyramid, and will in fact do so. That move is compression. And it will betray the ultimate return on invested capital (ROIC) on all this spend. ↗
- 2026-09-14 — Regarding “the science,”this is something you all need to know.
Most published science is fake or false
This is out of Stanford University and the author here John P. A. Ioannidis is one of the most-cited individuals on Earth. ↗
- 2026-09-14 — Let's all take a moment to understand how self-serving it is for OpenAI, Anthropic and other execs of big hyperscalers to talk of slowing things down.
1. LLMs are not AI and won't be AGI. There is nothing AI to slow down.
2. Competition is coming up fast, slowing benefits incumbents.
3. IPOs need hype & puffery; "we are so awesome it could become dangerous" is hype & puffery
4. Cover for real uncontrollable slowing growth as IPOs look to be pushed out ↗
- 2026-09-13 — Just a reminder. Plus there was the death spiral warning on BBBY.
Just trying to help out. ↗
- 2026-09-11 — When I ran money, investors wound ask me to share with them any alternative investments I like or can get behind. Now, I can share them with you.
They do not come along often. ↗
- 2026-09-11 — Shorting the Dollar by Drinking Free First Growth Bordeaux
Or, How a Lifetime Free supply of First Growth Bordeaux Requires Shorting the Dollar
Leg #1: A Bear Market in Wine
Fine wine as a category has been in a nearly 3-year bear market. Prices have tumbled roughly 25–30% on the Liv-ex indices from the October 2022 peak. Below, the Liv-ex 100 index, which tracks the price movement of 100 of the most sought-after wines.
For certain types of fine wine, such as here with the Liv-ex Bordeaux 500, the fall in prices has been more dramatic, and still has not bounced.
This is the deepest broad correction in the modern era, and a serious break from the rising prices of the last many decades. Wine pricing is in fact now below trend.
French Burgundy wines stumbled too. Burgundy, the Queen of Wines, is produced in smaller batches than Bordeaux, the King, but that was of little benefit as both lost their heads.
Kings and Queens and guillotines, sang Aerosmith
Aerosmith - Kings And Queens (Audio) https://youtu.be/b7jGGBxA3Yg?si=dBQUTXmuIEI9eBSf via @YouTube
The 2025-2026 recovery is still testing its legs. Broader wine price indices such as the Liv-ex 1000, which tracks 1,000 wines from across the world, are up only fractionally on the year, with Italian wines, perhaps overly slighted next to French wines over the last decade, leading the way.
Masseto, a Super Tuscan stalwart made entirely from Merlot, has seen its 2022 vintage move up about 10–15% year to date. However, more broadly, historically popular wines such as Burgundy and especially Bordeaux have barely moved.
The Price of Wine
An academic article published by the Journal of Financial Economics in November 2015, The Price of Wine, by Elroy Dimson, Peter Rousseau, and Christophe Spaenjers, is the current reference article on historical wine prices. The trio assembled 36,271 auction and dealer prices for the five First Growths from 1900 through 2012. They found a 5.3% annual real return before carrying costs and 4.1% net of estimated storage and insurance.
They also found, from 1900 to 2012, British equities produced a real return (after inflation) of 5.2%, which compares with 2.8% for British stamps, 2.4% for British art, 1.5% for British government bonds, and 0.9% for British Treasury bills.
The nominal return - after inflation - of course was much higher, but less interesting, except to those who market wine as an investment.
Now, fine wine beat the other collectibles and fixed-income assets in the study, and beat them all quite handily. Transaction costs and holding costs do vary across assets, but over 113 years, there is a lot of power in that study.
Still, fine wine is not an equities substitute, nor does it need to be. This is especially true of the First Growth Bordeaux, the top Burgundies, and the great Super Tuscan vintages.
One last chart drives home the timelines we are dealing with here while also putting to rest there is any reason to look at anything but fine wines.
Fig 1 from the paper also demonstrates the longevity of fine wine. Per their model, wine continues to rise, beating storage costs, for 30-40 years. I say 20, but 30-40 works even better, for one’s kids and grandkids. More on that later.
Why Wine? Why not Bourbon?
What do a bottle of 25-year Pappy Van Winkle bourbon and a McDonald’s hamburger have in common? Both are functionally identical to the next one of its kind.
Moreover, the Pappy evolves neither in its bottle nor in its glass; it never passes a drinking window and vintage years are not really a thing. This Pappy is 25 years old forever.
The fine bourbon primary market is concentrated among a handful of bourbon deities. It is deep, efficient, and nearly impossible to buy at discounts to market levels because there is no significant formal, novatable secondary market.
Wines are differentiated horizontally by region, producer, style. This is Pauillac against Pomerol against Vosne-Romanée against Bolgheri. Hundreds of estates, each with growing conditions, soil, irrigation, and wind that make its wines genuinely non-substitutable.
Wines are also differentiated vertically, within the château. This is by vintage, as the year’s weather writes a different wine’s storybook under the same label every year. Margaux 2015 and Margaux 2018 are distinct assets with distinct prices, scores, and maturity curves despite being the exact same vineyard and subject to the same winemaking techniques.
Plot the vertical against the horizontal to find thousands of points, each one a sparsely-traded market of its own. This is why a long spreadsheet of prices of the finest wines yields a mispriced Vega Sicilia or a magnum priced, atypically, lower per centiliter than standard sizes.
Furthermore, wine is on the clock. Unlike spirits, wine does change in the bottle. Each wine therefore tracks toward a drinking window that opens and closes, and every bottle of wine consumed anywhere on earth shrinks the inventory of that exact asset forever.
Supply destruction is both continuous and a bullish supply dynamic that spirits, watches, and art can never match.
Wines are consumption luxuries at their most expensive. Names like Lafite Rothschild, Latour, Margaux, Haut-Brion and Mouton Rothschild on the left bank, with Pétrus and Cheval Blanc as their right-bank peers. Romanée-Conti, Leroy, Roumier, and Rousseau in Burgundy. Sassicaia, Ornellaia, Masseto, and Solaia comprising the finest Super Tuscans, out of Tuscany.
In the restaurant, the prices of any of these will appear untouchable next to multitudes of more affordable wine.
Fine wines do not reside on your wrist for the next decades. They do not adorn ears or necks. They do not stand proud for years among the Joneses.
No, they are enjoyed very personally, with good friends and family, in the moment and then gone forever, after many years of patience and expense, save for the memories.
All three things make for something of a certainty in terms of fine wine appreciation, if bought well.
Right Now? Why Now?
...Read More...
...
Sipping Singularity
Yes, my method is admittedly more Boy Scout than Army Ranger, but the gist is significant negative correlation between the dollar index and bonded fine wine exists today and over the last 25 years.
However, what is clear is that in none of those years did the dollar face serious risk of debasement. The debt, AI and quantum-computing issues are relatively new in terms of both the severity and the proximity of the threat.
.... READ MORE...
To keep tariffs in the face of a weakening dollar would be too hard on American consumers. Politically there would be repercussions, and the tariffs would go away.
However, with a weaker currency, American consumers will face higher prices anyway.
Looming is the possibility that if the U.S. loses its ability to finance its deficits on the strength of privilege, it will have to
...READ MORE...
The 2010 Lafite Rothschild Autopsy: A Lesson in Price
I have done autopsies, so I feel qualified to do this. Recently, a dealer offered me Lafite Rothschild 2010. Parker rating 100, one of the legendary vintages, at a price below its 2011 release price. The pitch presented this as a bargain. It is actually a most instructive corpse, for those willing to do the dissection.
Fifteen years of holding a perfect-score First Growth Bordeaux returned less than nothing. But the wine never disappointed, as critics who retasted it have reconfirmed the perfect score. All the price appreciation happened at the front end, at the winery itself. Fifteen years of appreciation, within that original 2011 price.
How? It is not such a scary tale. No grim warning. Nothing from left field.
It's all about catching fireflies...
...READ MORE...
Read the full post at Michael Burry's Substack, Cassandra Unchained
https://michaeljburry.substack.com/p/shorting-the-dollar-by-drinking-free?r=4repfn&utm_campaign=post&utm_medium=web ↗
- 2026-09-09 — Hmm, OK?
“I am the house now, so when we intervene with the Japanese yen, I have pretty good insight into what the Japanese, what the Bank of Japan is going to do, what Japanese policymakers are going to do,” Bessent said at a Southern Methodist University event in Texas on Tuesday. “And you can bet against me if you want.”
I Am The Law https://youtu.be/i_Pga70f_YQ?is=G9M8P3y55QQgTTRk via @YouTube @Anthrax #Anthrax ↗
- 2026-09-05 — Breaking news to start a long weekend.
That’s some good fishy. ↗
- 2026-09-05 — There are more of these littering stock markets around the world than I can recall for a long, long time.
It is as if the cool kids just outright won. ↗
- 2026-09-05 — Birkenstocks are good for just about anything. $BIRK ↗
- 2026-09-04 — In Flames
Only for the Weak
https://youtu.be/q86owGQzaG0?si=3UxZSuEfIVl8cl6Q via @YouTube @InFlames_SWE $LULU ↗
- 2026-09-04 — Short Thoughts - Cassandra in Flames?https://michaeljburry.substack.com/p/short-thoughts-cassandra-in-flames?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web
In Flames Like a Phoenix; She Who has Fallen Shall Rise Again.
https://michaeljburry.substack.com/p/lululemon-athletica-lulu-where-is?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web $LULU #lululemon ↗
- 2026-09-03 — We are discussing Technical Analysis of all things over in the Cassandra Unchained Chat section. 244 replies since late yesterday. This will become a series, I can tell already.
Please join in if technical analysis with a value bent (or value-based investing with a technical bent) might be your thing.
https://open.substack.com/chat/posts/ac8f7c9d-3881-4866-a5cd-5dd09f8b16d9 ↗
- 2026-09-03 — Palantir: An Accounting
Palantir is back in the stratosphere.
The facts have not changed. Yes, FOMO is pushing companies to hire Palantir for now, but its competitive position gets more dire almost by the day.
Palantir filed its 10-K February 17th, and oh it had stories to tell.
Accounts Receivable & Days Sales Outstanding
Palantir’s accounts receivable (AR) is first up. The traditional metric for AR is Days Sales Outstanding (DSO), with higher DSO implying customers are taking their time to pay for Palantir’s services for one reason or another.
In 9 of the last 12 quarters, AR grew faster than revenue – a persistent pattern generally attached to nefarious tricks such as channel stuffing, aggressive revenue recognition, or extended payment terms used as sales concessions. For real subscription businesses, AR growth should track revenue growth closely. When AR is volatile or outgrows revenue, it means the company is booking sales faster than it is collecting cash.
I look to the filings for an explanation, and find something interesting.
2022 10-K (Dec 31, 2022): “No customer represented more than 10% of total AR.” AR was $258 million. This was pre-ChatGPT.
2024 10-K (Dec 31, 2024): A big change. Customer A = 26% of total AR, which stood at $575 million. But no customer is greater than 10% of revenue for the year. That means Customer A contributed less than $287 million in revenue (10% of revenues) but owed Palantir ~$150 million. Revenue could even be much less than $287 million, and closer to $150 million.
Now, Q4 2024 GAAP Net Income was $79 million. If Palantir had to write off 50% of its AR with Customer A, it essentially wipes out a full quarter of GAAP income. That was Palantir’s 5th consecutive quarter of GAAP profitability.
By the way, for that fourth quarter of 2024, Palantir added back about $282 million of SBC expense (to the $79 million GAAP income number) plus another $79.7 million SBC-related payroll taxes to arrive at its “adjusted” earnings number that it presents to Wall Street and Wall Street accepts. I don’t do that.
That adjusted number beat estimates. The GAAP number simply matched estimates.
The $282 million in SBC added back is calculated under ASC 718, which requires estimating the fair value of equity awards at the grant date, not at the date they vest or are exercised.
So, the adjusted number beat estimates on an SBC value assumption. As happens so often on Wall Street. Moreover, my research shows the shareholder cost of stock-based compensation well exceeds that estimate.
In any event, yes, in that circumstance, I am claiming ALL of that GAAP income disappears. This would also increase Customer B’s share of remaining AR since the denominator (total AR) just took a big hit. This is an AR death spiral as now Customer B (and other large customers) would be more concentrated with more bargaining power.
Of course, the write-off would come down the road, not during that same quarter.
That did not happen, yet. Which does not matter.
To get these articles in more timely fashion, and to get many times the number of articles, chat discussions, and impromptu AMAs, than Iwill publish here, subscribe over at my Substack. DIsclosure: I am short Palantir stock and own put options on Palantir stock. Use the link below to subscribe to avoid the substantial in-app fees charged by Apple and others. https://michaeljburry.substack.com/
When a customer A has 26% of your AR and knows you are a public company hyping your growth and adjusted earnings beats, Customer A acquires a tremendous amount of bargaining power because they know very well that you cannot afford a write-down.
So at contract renewal, the customer can demand price breaks, extended payment terms, and Palantir has to agree. And it doesn’t need to be 26% that triggers such bargaining power. It could be, for instance, 15% of AR.
Customers A’s AR grew from a 15% share to 26% over the course of 2024. Since there was no new disclosure of a customer over 10% of revenue, then either the customer was being invoiced ahead of delivery (channel stuffing) or the customer was not paying on time (or was on extended payment terms).
Neither option is good. One is accounting fraud, and one is a weakening business condition.
Q4 2025 10-K (filed February 17, 2026): Customer A still at 25% of AR, and the 5 year trend of rising DSO is not subtle. The average DSO has almost doubled from 35 days in 2020 to 66 days in 2025, and the year-end DSO is up 63% from 52 days to 85 days during that time.
This has coincided with the emergence of Customer A’s concentration in accounts receivable.
What we have here, it seems, is a customer holding the leverage because Palantir’s stock price is more fragile than the customer’s budget.
It is possible other customers are large as well and the same dynamic is playing out – a Customer B emerged at 11% of AR in the 3rd quarter of 2025, but slipped back under 10% at year’s end.
This is all about information asymmetry. Rising AR with growth yes, but rising DSO requires investigation.
Consider. Palantir’s government customers are more likely to have a higher AR share paired to a lower revenue share, and in 2018 and 2019, this was the case. Such customers had AR concentrations as high as 42%, with revenue concentrations only as high as 15%.
This asymmetric “small revenue share high AR share” customer fell back below 10% on both counts for the IPO. In 2021 and 2022, no customers, per the Form 10s, had more than 10% of AR or more than 10% of revenue.
Then,
2023, Customer A appears with 15% of AR.
2024, Customer A has 26% of AR.
2025, Customer A has 25% of AR.
At no point during these years does Customer A have more than 10% of revenue. This appears to be the same big government customer from before the IPO.
In the best case, Customer A today accounts for a whopping 7 months of DSO, and that is if the customer is only a hair under 10% of revenue.
If the AR/R ratio for this customer approaches the 2018/19 period’s 3x, DSO could be over 9 months. Or longer.
Any way I slice it, Palantir is losing either bargaining power or it is channel stuffing, or both. The former is a weak business position, and the latter a crime.
Do not laugh. That latter possibility is actually not so far out there. The pattern supports potential channel stuffing perhaps even more than a loss of bargaining position, and again they are not mutually exclusive.
For instance, big government Customer A could be trying to help Palantir stock. Perhaps some decision makers in government have owned and now own Palantir stock.
Maybe Customer A paid down its accounts receivable (from 42% of AR to under 10%) specifically to help the IPO and kept it that way for a few years after the IPO to help Palantir have a good launch as a public company.
If true, Company A would have demonstrated a willingness to help Palantir make its books look better.
That may sound good, but it is most certainly not.
That same big government Customer A of course could also “help Palantir” by allowing channel stuffing that puts Palantir over and above earnings and revenue targets more recently.
Again, Customer went from under 10% of AR to 26% of AR over the last three years. Government is known for taking time to pay, but 7 months, 9 months, 1 year, more than 1 year…and increasing.
As I said, consider.
The 10-K reveals that allowance for credit losses is “immaterial.”
Palantir has no reserves against any of this. This may not matter now, but I would wager this will be the state of things when it does.
To get these articles in more timely fashion, and to get many times the number of articles, chat discussions, and impromptu AMAs, than Iwill publish here, subscribe over at my Substack. DISCLOSURE: I am short Palantir stock and own put options on Palantir stock. Use the link below to subscribe to avoid the substantial in-app fees charged by Apple and others. https://michaeljburry.substack.com/
Deferred Revenue Patterns
Ok, here we go. We have nine companies with deferred revenue histograms by quarter over 5 years.
Do you see the pattern that is the kidney punch as to what Palantir is?
Here is the decoder.
Row 1: Palantir (GGGR), Adobe (RRRG), Salesforce (GRRG)
Row 2: ServiceNow (RRRG), CrowdStrike (RRRG), Intuit (RRRG)
Row 3: HubSpot (RRRG), Workday (RRRG), Accenture (GGGR)
Six out of seven SaaS companies show the same precise 1G/3R signature – every single year, without exception. Salesforce is GRRG because its fiscal year ends January 31, but otherwise fits the SaaS pattern. The only two panels that truly do not fit are Palantir and Accenture, the two consultants, and Palantir is now clearly in Accenture’s 3G/1R camp.
Bulls saying it deserves 70x, 90x, 110x revenues because it is a SaaS/Software company needs to explain why this is not.
I will explain why it is.
A true subscription software company like Salesforce or ServiceNow sells annual subscriptions. When a sale is made, a SaaS company books the full sale to deferred revenue – cash received, but not yet counted in sales. The SaaS company then recognizes that revenue and reduces deferred revenue proportionately through each of the 12 months of the contract.
Most SaaS companies also have a concentration of annual renewals in one quarter typically Q4 when most business renewal decisions are made.
So, for SaaS companies, deferred revenue spikes, then falls for three quarters. These are the changes charted above. Green and then three quarters of decline. Again, Salesforce has a January 31 fiscal year end which divides its deferred revenue bump and so is slightly different.
Consultancies do not sell annual subscriptions – they sell engagements such as a 3-month strategy project, a 6-month systems integration project, etc. In most quarters, consultants are ramping new engagements as well as executing existing engagements.
Deferred revenue is generally growing as the consultancy grows, but tends to fall during summer months when completions are high and new engagements are low.
Accenture has this dip in the 2nd quarter every year, and Palantir has this dip in the 3rd quarter every year – matching with the government’s fiscal year-end.
Palantir is clearly doing this, delivering labor and expertise on an ongoing basis, and billing ahead of completion. Its pattern has become more strongly like Accenture’s in recent years.
Now the uppercut. The Deferred Revenue/Revenue ratio at the top right corner of each panel shows every SaaS peer sits at 80-207% (207% being CrowdStrike), Accenture at 31%, and Palantir at 32%.
This is simply because subscription businesses have higher absolute levels of deferred revenue – actually the highest – for all the reasons described above.
Palantir is just not what it claims to be. It is growing so fast for the same reason World Wide Web consultants Razorfish and DiamondCluster grew so fast during the 1995-2000 data connectivity/online FOMO that hit C-suites across America. Everyone is scrambling.
Even if the scramble continues a few more years, the fall will be just as epic, or more so. And the current market cap will prove ephemeral.
To get these articles in more timely fashion, and to get many times the number of articles, chat discussions, and impromptu AMAs, than Iwill publish here, subscribe over at my Substack. DIsclosure: I am short Palantir stock and own put options on Palantir stock. Use the link below to subscribe to avoid the substantial in-app fees charged by Apple and others. https://michaeljburry.substack.com/
The $17.2 Million Mile High Club
Some have asked me to touch on Palantir as a political and unethical company. That can be a third rail, and I have tried to focus on the business model, the financials, the accounting, and the valuation.
Some aspects, however, fall within my purview here.
For instance, the Financial Times reported this morning on Palantir’s 10-K with a piece entitled, “Palantir CEO Alex Karp has his head in the clouds.”
Alexander Karp, the Company’s Chief Executive Officer, flies on non-commercial aircraft beneficially owned by him (the “Executive Aircraft”) for business and personal travel. During the years ended December 31, 2025 and 2024, the Company incurred expenses related to the use of the Executive Aircraft of $17.2 million and $7.7 million, respectively.
It’s quite the feat to spend $17.2mn in a year on “business and personal travel”, particularly when the jet’s not even a rental. Jefferies’ analyst Brent Thill runs the numbers:
Assuming use of a mid‑sized jet with an estimated operating cost of ~$7K per hour, this implies roughly 2,457 flight hours, or about 28% of the year spent in the air. Even under a more conservative assumption of a high‑end jet such as a G650 at an estimated ~$15K per hour, the $17.2M still equates to approximately 1,147 flight hours, or ~13% of the year. Notably, this $17.2M figure is more than double Karp’s executive aircraft expense in CY24 ($7.7M) and appears elevated relative to peers, with META CEO spending ~$1.8M and PANW [Palo Alto Networks] CEO spending ~$2.4M on private aircraft travel.
Profligate, egregious spending has long been linked to Palantir under Karp. So, in many ways this is not surprising. It is, however, one of the few concrete signs that we as readers of filings have as to the ethics and the nature of the CEO.
Who Pays Taxes? Not $PLTR
One would think it is notable Palantir lost so much money in the past that it paid zero federal taxes in 2025 on $1.6 billion pre-tax income, its putative third profitable year.
After all, many companies lose money for a time. When they return to profitability, taxes are sheltered because past net operating losses (NOLs) are carried forward to shelter taxes.
The really remarkable part though is that for Palantir, that is literally not even the half of it.
The stock-based compensation (SBC) is so enormous and stock appreciation so parabolic that it creates net operation losses (NOLs) that dwarf operating profits.
U.S. Federal NOLs ballooned to $9.0 billion, up from $5.5 billion, despite $1.6 billion in pre-tax GAAP income. Read that sentence again.
This would mean Palantir’s stock-based compensation-related tax deductions, and to an extent R&D amortization and other tax deductions, in 2025 alone were ~$5.1 billion. Now read that sentence again.
The party does not stop at the Federal level. State NOLs are at $4.8 billion, and UK NOLs are $1.8 billion.
The 2017 Tax Cuts & Jobs Act that in 2022 started requiring companies to capitalize R&D expense. This would normally be Godsend for a software company wanting to show higher current earnings. This is completely irrelevant to Palantir because its stock-based compensation-generated $9 billion in Federal NOLs.
You cannot make this stuff up.
How about this: in 2026, if Palantir’s stock continues to levitate, SBC tax deductions could be an additional $4-$6 billion. Year after that same thing.
Palantir still has 152 million options outstanding, and each one will produce a $135 deduction at a $145 stock price. That is over $20 billion in future SBC tax deductions adding to NOLs in the future, on top of this year’s $9 billion in Federal NOLs, if bulls are correct.
Shareholders are funding employee compensation through dilution.
The government is subsidizing Palantir through gargantuan tax shields.
The 10-K shows two founders and one board member have 10b5-1 trading plans adopted last quarter with the stock near highs that will have them all selling stock for most of the next year.
Palantir cancelled its 2023 $1 billion buyback authorization (with $860 million remaining) after buying back just $75 million worth of shares in 2025.
The stock price funds compensation, generates the tax-shield through NOLs, attracts talent and the hot $PLTR stock generates branding that subsidizes customer acquisition costs.
Cumulative GAAP earnings since the IPO is in the $2 billion range. The value of stock-based compensation transferred to employees during the period is likely greater than $9 billion market value at exercise.
With $PLTR, in so many ways, the stock price is the business model.
Former Forward Deployment Engineers Speak Out
I have been fortunate in that when I say I am looking into something, I get lots of offers to help.
I have to be careful, because I have learned when you engage crazies, they are very hard to disengage.
Every once in a while, the email or text is accompanied by bona fides I cannot ignore.
So it is that I found myself in the possession of an emailed copy of a very long Slack chat between former Palantirians regarding my post, “Palantir’s New Clothes: Foundry, AIP and the Failure of Reason.” The Slack thread is titled, “Palantir Alumni Thread – Reactions to Burry Short Report.” I was told all the contributors are former Palantir engineers, but I cannot confirm that.
They are not financial analysts, so some misconstrued certain claims, such as my RSU-related SBC argument, and R&D treatment. I will not repeat or address those here, but it is easy to understand. They also took issue with the “8 weeks for Foundry” claim. The wording I used is clear that development did not stop after 8 weeks, just that a working product was developed in that time. This matches Steinberger’s account in The Philosopher in the Valley, which I recommend everyone read.
Some other claims are a little confusing. Such as one former FDE saying that I refuse “to engage with the substance of what Foundry/AIP actually does, preferring instead to evaluate the idea that Foundry is AI (despite just being a wrapper around other vendors’ AI).”
Which both misunderstands my point and makes my point.
Another one says “I was nodding my head through some of this, but other parts are just inscrutable...Remember, Palantir R&D is not like normal software company R&D. Remember, the actual work is done not by Palantir’s Foundry but rather by the client’s installed base of third-party software. Palantir does not need to spend on R&D like real software companies.”
Which itself is an inscrutable statement because this is a fundamental claim of mine as well.
I have selected a number of quotes from the chat and listed them below. These are not in chronological order, and each quote is an excerpt. This is not the full conversation and does not represent continuity between any two quotes.
Person C [1:12 PM]: “he hits on a lot of points I do really agree with as a previous comm delta. Particularly about the services vs product aspect and how at least up to 2022 foundry had a really hard time sticking at an org and it was real hard to scale contracts”
Person B [7:40 AM]: “ pltr’s moat imo was the people. for a long time they were able to squeeze together a bunch of smart and hardworking folks.clients weren’t really paying for the software, they were paying for an elite team that will show up at 2 am and work 80h non-stop to fix their problems”
Person R [10:57 AM] “This is enjoyable reading despite the issues people have pointed out. It’s nice to hear the early days (my time there) portrayed accurately, though he misses that some of the incredible money burn came from very wasteful operations, not just the business model.” [MJB edit: I implied as much when I asked where all the money went and pointed out how Palantir throws money around.]
Person I [12:18 PM] Enjoying a cigar and on thumbs. Quick preview makes me chuckle. I don’t often find acknowledgement that P almost ran out of money multiple times.
Person K [7:56 AM] (a critic in general of my piece): “I fear the Palantir of today doesn’t have the ability to discover, much less react to, the very severe talent attrition. The mythical FDE quality of the before times is firmly in the past. A lot of very customer-visible mistakes are being made.”
Person R [11:09 AM]: “ as a former FDE from those days who saw a lot, he is correct to say that the IPO was lying to lump all those costs into R&D. You can’t capitalize first class flights to Sao Paolo.”
Person K [1:21 PM]: “I’m sure challenges could be made to whether every last team dinner and Thursday-Tuesday Gotham ‘colo’ is accurately considered R&D.”
Person F [11:01 AM] “You know what else has changed since ~2023? Anyone with a moral compass can’t justify working there anymore. Not an instant change, but it is getting worse every day since the original ICE fiasco during trump 1. Turns out that smart, competent people tend to be more liberal, more empathetic, more justice-based, and in favor of western and liberal values. That used to attract talent and now it is doing the opposite.”
Person B [7:40 AM]: “it’s been super noticeable since ~late 2023 with the quality of hires dropping and the most elite leaving or not even joining”
Person C [1:29 PM] “Palantir clearly jumped on the AI bandwagon in a way that for everyone in the company for the 5 years before must have felt like super whiplash. But I don’t begrudge that. Fair enough. And it’s likely true the foundry data foundation provides the optimal thing to use AI on, but damn rebranding Foundry to AIP was bandwagon to the max.
Person K [1:32 PM] “The whiplash is from the company line pre-ChatGPT being “AI is a distraction, data integration and ontology is what’s important”. Ironically, that remained true, while the marketing shifted dramatically.”
Overall, the Slack chat content did a lot to help my thesis.
Separately, a former FDE told me that “Foundry/Gotham is a lock in where Palantir then closes off the systems if a customer decides to leave.”
This has long been the story – Palantir’s stickiness with customers is not having a great product, but its obstruction of the exit ramp. Even so, as Person C said in the Chat, Foundry had a really hard time sticking with customer/organization. Getting customers to Scale up until 2022, at least, was very hard, he said.
My NDR data in the original post showed a decline into 2023, and I had pointed out that was indicative of a broken business model and high churn. That point is validated by these former FDEs.
Back in 2017, the NYPD alleged that Palantir was not cooperating with its attempts to discontinue its engagement. Eventually, per the Steinberger book, all police departments in the country left Palantir. Per a news report, at least one, in New Orleans, followed the NYPD in creating its own app for what Palantir did.
A former FDE also told me, “Foundry is not a perpetual license, you have to be trained to use it, even then you require heavy lift and continuous support.”
This matches the “80h non-stop” comment above that said clients were not paying for software, but for the FDEs/implementation consultants.
Overall, I have yet to hear from former Palantir engineers telling me my thesis is wrong.
I am left more convinced that Palantir does not earn the margins it says it does, does not have the earnings it says it does, and does not have the software subscription model it says it does.
Palantir is a consultant riding a bubble of AI FOMO demand, and will trade well below $100 billion market cap in my opinion. In the long run, revenues shrink, and it is likely acquired for relatively cheap. That is, after all, what happened to DiamondCluster and Razorfish.
Diamond was a struggling tech consultancy business, but pivoted hard as a World Wide Web consultant in 1997 and business exploded. Grew 60-70% CAGR through 1997-2000. The stock rose 300% in 1999. Market cap hit $2 billion. Got so cocky it bought a European operation Cluster for nearly $1 billion in cash and stock late in 2000. DiamondCluster was born,.
By the second half of 2001 DiamondCluster stock was down 88%, even though revenues initially only flattened. Revenues then fell, and the stock in late 2001 into 2002 traded at just 1x sales. In 2006, they sold the Cluster European operation, which they had bought for $1 billion in 2001, for $35 million. PwC bought the remainder for a de minimus amount, so small that it was never disclosed, in 2010.
Razorfish was a pure World Wide Web consultant, a comp to Palantir’s AIP. Razorfish IPO’d at $16sh in April 1999, doubled to $32 first day. The company hit a $4 billion market cap as sales grew 57% from 1999 to 2000 after growing 1,130% from 1998 to 1999. But sales then fell from $268 million to near zero in under two years. Razorfish was delisted in 2003.
Both companies had traded to about 20x sales at their peaks.
I do believe today’s AI Consultancy arc will play out in similar fashion. A rhyme, to some extent, if not a repetition. Palantir has climbed higher, and has further to fall.
Next up is Palantir: An Acccouning at the halfpoint of 2026. I will analyze the 10Qs and history in a similar manner to find patterns that most do not.
To get these articles in more timely fashion, and to get many times the number of articles I will publish here, subscribe over at my Substack.
Use the link below to subscribe to avoid the substantial in-app fees charged by Apple and others.
https://michaeljburry.substack.com/
Substack also includes a very lively and on-topic Chat area featuring a good number of impromptu AMAs.
DISCLOSURE: I AM SHORT PALANTIR STOCK AND I OWN PUT OPTIONS ON PALANTIR AS WELL. ↗
- 2026-09-03 — I agree, investors need to read this.
https://www.thestreet.com/investing/stocks/michael-burry-sends-another-nvidia-stock-verdict-to-investors ↗
- 2026-09-03 — Short Thoughts July 24, 2026 - Offshore Insurers, Meet the Hyperscalers
Also, Andrew Granato of the University of Texas and Pranjal Drall in a 65 page academic paper take down the private credit/private equity/insurance shell game
Michael Burry
Jul 24, 2026
This will indeed be a short thought, because I want the focus to be on this paper, “Private Credit’s State Backstop: How Private Equity Socializes Risk Through Insurers,” written by Andrew Granato and Prangal Drall.
Download📷Private Credit State Backstop Granago Drall July 21 20261.38MB ∙ PDF file
Andrew Granato is a Google Scholar and Assistant Professor at the Univeristy of Texas School of Law. He is on X, handle agranato42. Andrew has B.A. Economics from Stanford. He also posses both a J.D from Yale Law School and a Ph.D in Financial Economics from Yale School of Management.
Prangal Drall is a doctoral fellow at Yale, where is pursuing both a J.D. and a Ph.D as well.
These may be my two favorite people in the world this weekend.
That does not matter too much. What matters is these two gents are qualified to write this paper on a subject that I have been trying to bring to the fore. For instance here,
The Heretic’s Guide to AI’s Stars Part III: Tracepalooza & the Bezzle
Again, the focus is on the paper, “Private Credit’s State Backstop: How Private Equity Socializes Risk Through Insurers.
The table of contents promises great things. They often do.
Exciting stuff, especially those sections III and IV.
I quote from the paper’s abstract.
“Private equity (PE) firms have acquired large life insurers and loaded their balance sheets with private credit assets that are opaque and difficult for regulators to value...when a life insurer becomes insolvent, state-based guaranty funds protect insurance policyholders by "assessing" surviving insurers to cover the shortfall. In most states, such outlays are fully creditable against state premium taxes over time..PE-owned life insurers reflect a structural transformation in which an insurer supports a broader asset-management business that is designed to extract value upfront and impose losses on others. PE firms exploit this regulatory regime by pairing life insurers with private credit to capture value from both sides.
You have heard shades of this from me before, and I have been building to it in Heretic’s Guide Part IV, not yet published.
Let’s start with how it can all end. Page 30 of the paper brings up an important point about the end game, if and when the insurers go insolvent.
For the Rest of The Story , please visit Cassandra Unchained on Substack. ↗
- 2026-09-03 — I'll take it.
https://www.thestreet.com/investing/stocks/michael-burry-trims-veeva-veev-cuts-paypal-pypl ↗
- 2026-09-02 — No better death metal song. @TheEntombed
https://www.youtube.com/watch?v=17S-d5S7Ms8&list=RDL-lvai3M7K0&index=2 ↗
- 2026-09-01 — Never before, I say. ↗
- 2026-08-28 — Surprise to who? Not CU.
Meituan Has Surprise Profit After China Delivery Fight Eases (2)
Summary by Bloomberg AI
▪Meituan reported a surprise quarterly profit after competition in China’s quick delivery market eased and the company slowed its overseas expansion.
▪Operating profit rose to 2.69 billion yuan for the three months through June, and revenue climbed 14% to 104.6 billion yuan, also topping estimates.
▪The company's domestic delivery business returned to year-on-year growth "primarily due to our disciplined spending on incentives", according to Meituan.
By Luz Ding
08/28/2026 05:33:03 [BN]
(Bloomberg) -- Meituan reported a surprise quarterly profit after competition in China’s cutthroat quick delivery market eased and the company slowed its overseas expansion.
Operating profit rose 11-fold to 2.69 billion yuan ($400 million) for the three months through June, the company said Friday. Analysts estimated a loss of 846 million yuan on average.
Revenue climbed 14% to 104.6 billion yuan, also topping estimates. Net income reached 2.16 billion yuan, after three quarters of losses. ↗
- 2026-08-27 — Whistling past the graveyard ↗
- 2026-08-27 — Addressing the issues of the day in real-time here.
https://michaeljburry.substack.com/p/trading-post-august-27-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web
Michael Burry's Trading Post August 27, 2026 ↗
- 2026-08-27 — The institutional imperative is powerful and rarely fails in its subversion of thought. ↗
- 2026-08-23 — Bullwhip will sting. ↗
- 2026-08-23 — A few months back I flipped my $BABA stock into a large $JD position. I will not flip any of that back to #Alibaba, as issuing shares is now its new paradigm. It would have to fall by half from here for me to look at if again. ↗
- 2026-08-21 — Trading Post & Short Thoughts to the Max, August 18, 19, 20, 2026 -
https://michaeljburry.substack.com/p/trading-post-and-short-thoughts-to?r=4repfn&utm_campaign=post&utm_medium=web
The AI Buildout, Small Language Models, Intelligence per Watt, & Buying and Shorting 24x5
Things are getting concentrated around here. ↗
- 2026-08-20 — LIttle Wshes is a highly rated 501(c)(3) non-profit public charity that grants the immediate and ongoing wishes of chronically and critically ill hospitalized children.
https://www.littlewishes.org/
January 2026 Cassandra Unchained Charity of the Month
https://michaeljburry.substack.com/p/january-2026-cassandra-unchained?r=4repfn&utm_campaign=post-expanded-share&utm_medium=post%20viewer
What both breaks and lifts my heart is the brave spirit of these kids as they face, again and again, trying tests, procedures, surgeries and other treatments. It seems no matter what, in quieter and more lucid times, they stand ready to smile. ↗
- 2026-08-20 — Who wants to be the next George Soros?
*BESSENT: BUYBACKS COULD BE BIGGER THAN THE $4B WE ANNOUNCED
*BESSENT: US YIELDS DON'T REFLECT UNDERLYING FUNDAMENTALS
*BESSENT: LIQUIDITY IN 30-YEAR POINT ESPECIALLY POOR ↗
- 2026-08-20 — R & W.
https://arxiv.org/html/2511.07885v5 ↗
- 2026-08-19 — Highlighting our December 2025 Cassandra Unchained Charity of the Month
https://togethercalifornia.org/
Together California is an inspiring public charity working to dramatically improve the experience and outcomes for children in the foster care system.
https://michaeljburry.substack.com/p/december-2025-cassandra-unchained?r=4repfn&utm_campaign=post-expanded-share&utm_medium=post%20viewer ↗
- 2026-08-18 — This is serious competition for NVDA.
The startup says it took just 44 days after getting its test chips back from Taiwan Semiconductor Manufacturing to have them up and running inference workloads—the computing processes that allow AI models to respond to user queries—a process that usually takes six months or more.
When Etched needed help designing servers and racks to integrate its chips into one big system, the company in 2024 recruited Brian Loiler, a top systems engineer who spent nearly 23 years at Nvidia. Loiler has since recruited to Etched about a dozen engineers from Nvidia who, in some cases, turned down attractive counteroffers, he said.
https://www.wsj.com/tech/ai/a-21-billion-kids-in-chips-startup-is-scooping-up-nvidia-talent-4d099f12?st=K3FaWU ↗
- 2026-08-18 — Acres of Diamonds https://acresofdiamonds.org/
Cassandra Unchained's August 2026 Charity of the Month
https://michaeljburry.substack.com/p/acres-of-diamonds?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web
Acres of Diamonds is a 501(c)(3) organization that sets out to ensure that every woman and child experiencing homelessness that comes to Acres receives the opportunity to rewrite their story. With families experiencing generations of poverty, abuse, neglect and addiction, Acres of Diamonds aims to provide holistic, long-lasting solutions to help address the problems.
5% of gross revenue from Cassandra Unchained during the month of August will be donated to Acres of Diamonds to further their mission and the good work they do. ↗
- 2026-08-18 — Peak Sepultura.
https://youtu.be/8daMimkWAHA?si=qdpcFjb1i7tcrZSv ↗
- 2026-08-17 — Well, you could have heard it first, months ago, 2025 even
Like compression’s threat now, to be covered in 2027 by @WSJ and @CNBC ↗
- 2026-08-13 — Joe Satriani killing it, to no one's surprise. That's a Dragon amp, built by 3rd Power and Joe to nail Eddie Van Halen's Live Without a Net tone. 100%.
Panama (Van Halen) - Sammy Hagar, Joe Satriani, Michael Anthony, Jason B... https://youtu.be/PQVmU1CVqDU?si=cw0sD0fXhdXcnIvo via @YouTube ↗
- 2026-08-09 — Now for something different.
Parallel Structure: the Pickup in Eddie Van Halen’s Frankenstrat on #vanhalen’s first album in 1977
https://open.substack.com/pub/michaeljburry/p/parallel-structure-the-pickup-in?r=4repfn&utm_medium=ios
I wound so many pickups in so many different ways but this pickup/guitar configuration convinced me.#evh ↗
- 2026-08-08 — (media) ↗
- 2026-08-07 — Kakashii is a must-subscribe. Bull or bear, his research must-read, must-think, must-act material. ↗
- 2026-08-05 — My full quote. ↗
- 2026-08-04 — I Helped Run Lululemon. Companies Need to Stop Kidding Themselves About A.I. https://www.nytimes.com/2026/08/03/opinion/ai-hype-tech-layoffs.html?smid=tw-share ↗
- 2026-08-02 — Do not forget your weekend's mandatory read.
What happened last week explained.
Foundations: Market Structure, Volatility Targeting, Pod Shops & Other Gremlins
https://open.substack.com/pub/michaeljburry/p/foundations-market-structure-and?r=4repfn&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
The Type of Read that Grants the Reader not only Understanding but Air Superiority ↗
- 2026-08-02 — Genius is found at the cusp of disciplines.
https://youtu.be/WkXAAHDIN7U?si=mABOOPnJrbrkj6g4 ↗
- 2026-07-31 — Foundations: Market Structure & What The Heck Just Happened
The Type of Read that Grants the Reader not only Understanding but Air Superiority
https://open.substack.com/pub/michaeljburry/p/foundations-market-structure-and?r=4repfn&utm_campaign=post&utm_medium=web
For your weekend reading, studying, tripping, whathaveyou. ↗
- 2026-07-31 — The real news is OpenAI slashing and burning its prices to prepare for this.
#DeepSeek releases beta version of V4 models as AI price war heats up. https://asia.nikkei.com/business/technology/artificial-intelligence/deepseek-releases-beta-version-of-v4-models-as-ai-price-war-heats-up ↗
- 2026-07-30 — New Trading Post today.
https://open.substack.com/pub/michaeljburry/p/trading-post-july-30-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-07-30 — This is called gaining experience. ↗
- 2026-07-30 — What’s going on today?https://substack.com/chat/6819723/post/46d5ff49-5e6a-450a-9f3e-b4ca20b597be
That Momenum Pair Unwind was the most dramatic ever, incluidng 26 years ago. So it was due for a bounce the other way. Bear markets have the most dramatic rallies. Especially at the beginning.
I have been explaining and discusssing in the Chat area, My Thought Pad and the companion thread discussing the Thought Pad. ↗
- 2026-07-29 — Some posts are timeless. Others become timeless because they were well-timed.
About That Boy Who Cried Wolf
https://open.substack.com/pub/michaeljburry/p/short-thoughts-may-10-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-07-29 — Phosphorous
https://www.youtube.com/watch?v=Rcmi6_fMhyw ↗
- 2026-07-29 — There is a reason $NVDA’s 5 year credit default swaps are going parabolic. All this overreaching by #nvda to push the circular spending to biblical proportions . ↗
- 2026-07-28 — Foundations: The Psychology of Investing in the Information Age
CNBC’s Warren Buffett Archive, AI and Decision-Making
https://open.substack.com/pub/michaeljburry/p/foundations-cnbcs-warren-buffett?r=4repfn&utm_campaign=post&utm_medium=web
"Excitedly “cramming” all at once will only mean long-term memory is not attained for the content one’s prefrontal cortex so urgently consumed.
Worse, that same prefrontal cortex becomes overloaded with information that cannot be properly processed. This produces cognitive fatigue, affecting the rest of your day in negative ways.
Yes, cramming for one thing makes one dumber for other things. What’s worse, it gives one more ADHD, which no one today needs.
Most of us learned this intuitively during college. Or, are realizing it now." ↗
- 2026-07-28 — Evil incarnate. ↗
- 2026-07-27 — Ta-da! ↗
- 2026-07-27 — The market has voted and the results are clear. ↗
- 2026-07-27 — Shades of John Chambers. ↗
- 2026-07-27 — Around and around we go.
Nvidia to guarantee $200 billion of ChatGPT’s spending on $NVDA chips. ↗
- 2026-07-25 — Don’t forget your weekend reading!
This is important information of which all should be aware.
…A loan turns out to be worth less than was paid for it, and the majority of the loss, depending on the leverage, is born by the guaranty system. That guaranty system ends with the taxpayer…
Read all about it!
https://open.substack.com/pub/michaeljburry/p/short-thoughts-june-24-2026-offshore?r=4repfn&utm_medium=ios #apolloglobalmanagement $APO #MetLife ↗
- 2026-07-25 — Those assets were never meant to mix, so fast-growing is not really what we want. http://x.com/i/article/2080823310816886784 ↗
- 2026-07-24 — Everyone should read this. Everyone. ↗
- 2026-07-24 — The last can private equity will kick down the road
"Private equity (PE) firms have acquired large life insurers and loaded their balance sheets with private credit assets that are opaque and difficult for regulators to value...when a life insurer becomes insolvent, state-based guaranty funds protect insurance policyholders by "assessing" surviving insurers to cover the shortfall. In most states, such outlays are fully creditable against state premium taxes over time..PE-owned life insurers reflect a structural transformation in which an insurer supports a broader asset-management business that is designed to extract value upfront and impose losses on others. PE firms exploit this regulatory regime by pairing life insurers with private credit to capture value from both sides" #privateequity #privatecredit #insurance
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7152239 ↗
- 2026-07-24 — Trading Post July 24, 2026 - 4 Shorts, 3 Longs
No frills trading update on a busy day, a word on Palantir, and how to find your peace on top a pile of gold.
https://open.substack.com/pub/michaeljburry/p/trading-post-july-24-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=post%20viewer #nvidia #qqq #patience ↗
- 2026-07-24 — When whine works.
(Bloomberg) ---- A US government agency has withdrawn a bid to improve a military intelligence system that is at the heart of lethal targeting following a protest from Palantir Technologies Inc., according to people familiar with the matter.
In its protest, Palantir contended that the DIA was “impermissibly seeking a developmental solution when a commercial solution is already available,”
https://www.bloomberg.com/news/articles/2026-07-24/after-palantir-beef-us-withdraws-bid-to-improve-military-intel ↗
- 2026-07-24 — Patience is the single most difficult skill to master for investing well. Too, the most impactful.
Position sizing trails patience. The more patience, the largera a position can be.
One must not mistake patience for foolhardiness.
A hardy fool is just a hardy fool. ↗
- 2026-07-23 — To Ride, To Shoot Straight & To Speak the Truth
Chainsaw Guitars
Chainsaw Vocals
RIP LG
https://youtu.be/tMMYLd8bq2s?si=RgJmz5FFJxkhFrZx&t=911 ↗
- 2026-07-23 — For those interested I have a running a chat room Thought Pad on Substack.
I post thoughts, charts, news/analysis day and night. There is a companion thread with an active discussion of my Thought Pad postings as well as other AMA type questions.
https://open.substack.com/chat/posts/46d5ff49-5e6a-450a-9f3e-b4ca20b597be?r=4repfn&utm_medium=ios&utm_source=share ↗
- 2026-07-23 — Hmmm…compression is not about open source per se. Compression is about the limits of knowledge, the use of small parameter proprietary models next to open source large and small parameter models. The idea that humans only have so much knowledge, that the use of that knowledge is limited and often redundant, that these models will become so compressed they will be run on edge compute. It would mean the buildout becomes overbuilt, that AI models are commoditized and customized. It rejects the idea of agent-to agent as of limited practical use and too consumptive of capital and other resources to be anything but carefully executed. ↗
- 2026-07-23 — Watch the long bonds. Treasuries pressured by AI's debt explosion, rising inflation vol, (per the great chart from @Bloomberg's Simon White). Basis Trade's shaky state, #oil back near 100, many factors, many charts. Not sure how much longer PE and PC can hold their breath. ↗
- 2026-07-23 — Trading Post July 23, 2026 - 3 Stock Buys & the BIS Weighs in on AI's Rampant Circular Financing
Taking the Momentum Pair Trade challenge by shorting the Pods
https://open.substack.com/pub/michaeljburry/p/trading-post-july-23-2026-3-stock?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-07-23 — There is a lot here.
Palantir’s New Clothes: Foundry, AIP, & the Failure of Reason https://open.substack.com/pub/michaeljburry/p/palantirs-new-clothes-foundry-aip?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web $pltr #palantir ↗
- 2026-07-22 — Incredible rare footage of Randy Rhoads absolutely killing with his Sandoval Polka Dot guitar with the converted and heavily modded Danoelectro neck.
https://youtu.be/VMACc_dTABg?si=witwOLpW4IvAIrkt&t=293 ↗
- 2026-07-22 — This is one of the better videos showing Randy Rhoads playing the Sandoval Polka Dot V with the converted Danelectro neck.
https://youtu.be/VMACc_dTABg?si=QGOcvm1N89eIVSbd&t=293 ↗
- 2026-07-22 — Who needs humanoid robots? The ChatGPT moment was free and instantly made cheating in school a literal no-brainer. Nothing can spread as fast or make more of an impact than turning off the brains of all students across the world in one instant.
https://asia.nikkei.com/spotlight/nikkei-forum/global-digital-summit/global-digital-summit-2026/china-s-unitree-says-gpt-moment-for-robots-remains-years-away ↗
- 2026-07-22 — Could be smart phones castrated society. ↗
- 2026-07-21 — PALO ALTO, California -- Hidden debt at U.S. tech giants swelled eightfold in roughly four years to an estimated $1.65 trillion as artificial intelligence investments ballooned, a Nikkei study shows, exceeding actual debt and making it tougher for investors to assess risk.
Nikkei examined recent financial statements and other materials from Google owner Alphabet, Microsoft, Amazon, Meta and Oracle. The four companies aside from Oracle are scheduled to announce their second quarter earnings from Wednesday, meaning the figures may increase further.
The five companies' hidden debt, which does not appear on balance sheets, totaled $1.65 trillion in the most recent quarter, exceeding the roughly $1.35 trillion in debt reflected on their balance sheets. The data includes some estimates. ↗
- 2026-07-21 — NASDAQ 100 futures up 282, but yep. ↗
- 2026-07-20 — The whole thing I think it's sic.
https://youtu.be/IDwaMfOyGj4?si=i8HWEqkKlUomwkQr&t=164 ↗
- 2026-07-20 — An urgent note on electricity's grand arrival and its parallels with the current #AI buildout. And, of course, much more.
Enjoy the show! Light the World!
https://open.substack.com/pub/michaeljburry/p/the-blessed-fraud-recurrence?r=4repfn&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true $NVDA $CRWV $ORCL ↗
- 2026-07-19 — 95% of investors likely have no idea what they really own. Let me re-phrase that. 95% of investors like to have no real idea of what they own. ↗
- 2026-07-19 — Alibaba launches a preview of Qwen3.8 Max, describing it as comparable to the frontier models of all but Anthropic. ↗
- 2026-07-19 — We have done it before, substituting a societal hallucination for a working system and excelling inside that hallucination to the detriment of real world stability and progress. ↗
- 2026-07-18 — For a weekend deep think, consider this. ↗
- 2026-07-17 — Trading Post July 17, 2026 Plus Netflix & What's Up with the VIX
Added to one stock position, cut one put position in half, and added to another put position
https://open.substack.com/pub/michaeljburry/p/trading-post-july-17-2026-plus-netflix?r=4repfn&utm_medium=ios ↗
- 2026-07-17 — It is a particularly good time to look to Hong Kong for cheap stocks that should do well as the shine comes off Korea, Japan & the Soxx.
https://open.substack.com/pub/michaeljburry/p/hong-kong-stocks-structure-and-strategy?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-07-16 — My whole life, none have pressured me more than myself. ↗
- 2026-07-16 — Short Thoughts (but long post) July 15, 2026 PYPL/IBM/HCA/More
Trading Post Plus Commentary on $PYPL Buyout Offer, $IBM and $HCA Swoons, Ben Graham & #GrowthInvesting
https://michaeljburry.substack.com/p/short-thoughts-july-15-2026-pyplibmhcamore?r=4repfn ↗
- 2026-07-16 — Must understand for Fannie and Freddie investors. $FNMA $FMCC - this EO report is not necessarily public and may already be in the administration’s hands.
Order 14393 Promoting Access to Mortgage Credit was signed March 13, 2026. This EO gave the FHFA director 120 days to submit a report on the efficiency of national housing finance markets, recommendations on regulations or legislation. That is due right around this July 11th weekend.
https://open.substack.com/pub/michaeljburry/p/washington-goes-to-fannie-mae-and?r=4repfn&utm_medium=ios ↗
- 2026-07-15 — A helping hand is all ↗
- 2026-07-15 — Short Thoughts July 15, 2026 PYPL/IBM/HCA/More
Trading Postette Plus Commentary on PayPal Buyout Offer, IBM and HCA Swoons
https://open.substack.com/pub/michaeljburry/p/short-thoughts-july-15-2026-pyplibmhcamore?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-07-15 — Wheyew!
https://youtu.be/3lAtSEcojiE?si=jYjxTH42lhNSMI4Y&t=775 ↗
- 2026-07-14 — What's happening. ↗
- 2026-07-13 — Parallel Structure: The Pickup in Eddie Van Halen's Frankenstrat on Van Halen's First Album
Below is a picture of Eddie Van Halen in December 1977, with a decent picture of the Frankenstrat in an early incarnation. Maybe the incarnation, and pickup, that he used on Van Halen's first album recorded three months earlier.
All of the following is how it went down, logically, according to my studies, my pickup winding, guitar building and amp repair experience, and how my brain works.
The pickup was a Gibson PAF pickup, perhaps from an early Gibson ES-335 that Eddie ruined experimenting with it.
The pickup broke before the first album, meaning it stopped making sound earlier in 1977. He tested the coils, and one was dead.
Instead of throwing the pickup away, he decided to fix the pickup and make it hotter, with more power, for the album. He used other pickups in the guitar while he worked on that one PAF pickup.
Since he had wax potted the pickup, the pickup could not easily be unwound to find the break in the coil wire.
Maybe he heard the Duncan Distortion or some other newer pickup was wound with thinner wire, and maybe he liked that idea because he could put more winds on it for, in his view, more power, or edge or cut, or something. Just more.
So, he completely rewound the dead coil by hand, but stuffed as much 43awg Plain Enamel wire as he could fit on the coil. This was thinner wire than the 42awg Plain Enamel wire used in vintage Gibson PAF pickups, including the unbroken PAF (slug) coil for Eddie's pickup.
Then Eddie unintentionally installed the repaired coil in a manner that put the coils in parallel rather than in series. Series is typical for humbucker pickups.
Parallel coils are still humbucking but have a more wiry, less full, more detailed and touch-sensitive sound than the Series orientation. I also hear a lifted shift in the mids that is oddly more textured than when in series.
For a novice working to put back together an original Gibson PAF, it is remarkably easy to accidentally solder the finish of the new coil directly to the baseplate ground along with the other coil's start, accidentally creating a parallel circuit.
Think about putting this back together. No internet. No AI. Just you and the mess you made.
In any event, that asymmetrically wound parallel-circuit modified PAF pickup carried him through the years, in and out of guitars and it may have changed some.
On that first album, the tone was different, and in the recording of a couple live shows played in 1977 around the time of the recording, that same tone was in evidence.
When Suhr measured it in the late 1980s and said it was such a low reading, but it wasn't a single coil sound, they thought that one coil had shorted.
Well if it really shorted, the coil would likely not work at all, but that was not the case. It could also naturally short in a way that makes the circuit parallel rather than series.
If the pickup had survived with no further modifications from 1977to the late 1980s, I would think what they really found was the humbucker circuit was in parallel.
This, from way back when Eddie knew very little about pickups, would explain the extremely low DC resistance reading.
"Nobody taught me how to do guitar work: I learned by trial and error. I have messed up a lot of good guitars that way, but now I know what I'm doing, and I can do whatever I want to get them the way I want them. "- Jas Obrecht November 1978 EVH Interview
So I wound a pickup according to the formula.
I took the cover off to put it in the guitar, but you can see my scratch, screw coil at 7.66K DCR with 43awg wire and slug coil at 4.36K with 42awg. Total DCR 12.16 at a hot garage workstation calmed down to 11.9K.
I bolted that asymmetrically wound pickup at an EVH slant in an ash body flatsawn two-piece maple neck strat-type guitar with vintage tremolo.
I did that with 4 conductor wire so I can put the pickup on a series/parallel switch.
Yes, I do believe he had two knobs on his guitar at the time of VH1, and a tone pot was in the circuit. I put that on a switch too - tone pot in, tone pot out. Tone pot out just did not sound right up high.
Here is the guitar with two white strat guitar control knobs in August 1977, just weeks before Eddie entered the studio to record VH1.
Also, it is easier to leave the tone knob in the circuit if the novice does not really understand the circuit. Perhaps this was the case with Eddie.
Later, he had a tone knob hidden in the control cavity of his famous original Bumblebee guitar.
I wound so many pickups in so many different ways but this pickup/guitar configuration convinced me that the parallel asymmetrically wound pickup described above is the right first album sound, at least where Eddie used the Franky.
When I switch from series to paralell with that asymmetric pickup while playing very loud through a Marshall style MGL50 amp and 4x12 with 1967-1968 20W G12M speakers, the change in tone is very apparent and very recognizable.
Full pickup specs on the label.
The vintage Gibson magnet had degaussed and gotten a little weaker over time for some reason.
November 1978, little over a year after the recording of the first album, Jas Obrecht's famously interview Edde for his story, "Heavy-Metal Guitarist form California Hits the Charts at Age 21."
"I've also recently bought a Charvel Explorer-shaped body and put a Danelectro neck on it and an old Gibson PAF pickup. And I also found a 1952 gold-top Les Paul. It's not completely original-it's got a regular stud tailpiece in it, and a Tune-o-matic bridge. I have rewound Gibson PAF pickups in it, too. I use a Les Paul for the end of the set because my Charvel is usually out of tune, and the Les Paul's sound is a little fatter. " - Eddie Van Halen
Eddie mentions rewound vintage Gibson PAF pickups in this and other interviews. That make no sense if he means fully rewound. Taking all the magnetic wire out of the pickup, then rewinding with new wire is actually insane.
There is no reason to obtain the PAF pickup in the first place with such a plan, as the sought-after tone is mostly in that wire, how it was made, and how it was originally wound.
But Eddie really was looking out for more old PAF pickups, chasing some sound. Jas Obrecht interviewed Eddie Van Halen again in 1979, as below.
Jas Obrecht: What other pickups did you try?
Eddie Van Halen: That’s about it. I’d do anything to get an old P.A.F.
Jas Obrecht: An old Gibson P.A.F.?
Eddie Van Halen: Yeah, they’re the best.
Jas Obrecht: I’ll keep my eye to the ground. I run across them every once in a while.
Eddie Van Halen: They go for 100, 200 bucks apiece, but that’s what I use, that’s what I like. A lot of people don’t like ’em. See, with my setup, it’s matched. If I play my guitar through someone else’s setup, it won’t sound right. If I use someone else’s guitar through my setup, it won’t sound right.
So, he acquired PAFs, keeping the slug bobbin stock and the rest of the pick-up stock but rewinding the screw coil hotter with 43awg wire.
Maybe he liked the sound of that first one he did, so tried to do it to other PAFs, with varying results since Gibson PAFs were not all standardized well. Maybe his asymmetry mod turned out great, and he knew he would need more PAFs to modify to keep his sound and his secret. I'm not sure.
I had built 4 EVH-style guitars for testing different pickups spec'd according to various EVH pickup theories.
For instance the old Dimarzio Distortion pickup in the black one has a vintage short Alnico 2 PAF magnet and also had one coil rewound.
The short vintage Gibson Alnico 2 magnet is used there because on the real bumblebee, this picture shows a short-length magnet. That is a unique sound too, but "Unchained" I think came from something else. I get great "Unchained" tones from my R0 Les Paul.
In any event here are the four guitars that swapped pickups and necks during my time doing this.
The Bumblebee-ish guitar has a vintage 1960 Danelectro neck with a thick steel bar running its length to keep it straight. The neck therefore is not adjustable. Luckily it is perfectly straight after all these years, and never moves.
This neck has a flat 14" radius and a comfortable back profile, and was popular with both Eddie Van Halen and Randy Rhoads.
Here is Eddie with one of his stripped down Stratocaster guitars with the Danelectro neck in late 1978.
The neck was also seen in the studio on probably the same stripped down Stratocaster body during the recording of Van Halen's second album in late 1978.
Rumor has it that it was used to paly the intro on Women in Love.
Lesser known is that the famous Randy Rhoads Sandoval Polka Dot guitar's neck had a Danelectro neck that was heavily modified with bow tie inlays and a rebuilt headstock. The heel was also modified to be used as a glued-in set neck instead of a bolt-on neck. You can see that very flat 14" radius in this picture of the GGOAT playing his Polka Dot.
If you are thinking, dang that must be a special neck, well that is what I though some years back and is why I sought one out to try on a guitar.
It really is a great-sounding neck, once wrangled into the neck pocket, which had to be modified just a bit due to the neck's 25" scale instead of the Fender/Charvel standard 25 1/2". It sounds very clear on all strings up and down with no dead spots. In fact, the neck rings a good long time.
I have also had an EVH Wolfgang guitar, Peavey EVH guitars, EBMM EVH guitars, and I had the original signed aged Frankenstrat pickup (1 of 400) signed by EVH himself.
I also tried aftermarket EVH-style pickups, and the Duncan 78 model.
All great pickups, and the relatively small signature guitars generally were very easy to play. Nothing too special. And not parallel.
I purchased a 1977 Ibanez Destroyer with stock pickups to try that out. Some think EVH used a guitar like this, maybe with stock Super 77 pickups, which had Alnico 8 magnets. Or, maybe not.
The Destroyer had a distinctive sound, and was not such an easy player.
I thought this Hamer Korina (not Sen, of course) Standard was a better player with far better sound.
Playing around with 70s style pickups Eddie might have used, I wound my own version of the Super 77 pickup, as well as my own version of the Mighty Mite 1400.
I found a vintage 1970s Alembic Hot Rod kit on the theory that Eddie would have used it somewhere. It is mostly just a giant ceramic magnet. I used it in the MM1400 clone, which sounds very raw, wild.
This pickup is in the blonde transparent strat with the cutaway pickguard. I think it sounds somewhat close to the VH1 sound as well, even though it is wired in series.
All things considered, the real VH1 sound is found in the pickup as I originally described. Asymmetric. Parallel.
Plus, I think the guitar is supposed to be homemade and imperfect.
The guitar should fight back a bit if it wants to do EVH. The 1st string even falling off the fretboard, as with those EBMM EVHs. Below, my favorite one, a hardtail, with that high E just getting ready to jump.
For VH1, imperfect brass nut, strange intonation, action a little high.
To my ear, the fight is part of it.
This early look at the Frankenstrat neck - those strings have very little break angle over the nut due to tall tuner posts and no string tree, all to make the strings come back in tune better over that brass nut on bends and trem use.
That's a fight.
I cannot imagine Randy's guitars played so great either with all the neck experimentation going on back then.
So, here's to the fight, and the fun, of it all.
This was a hobby (to de-stress) for me for a couple decades. I have passed it by and am now focused on my writing.
I still am building amps and tweaking the ones I have. That is another rabbit hole.
For now, I am trying to get down what I can about what I have done, my whole EVH pickup saga was a very long, unhealthy and drawn out affair. It ended with the parallel switch on the asymmetric pickup.
And that is that. No, no clips on X. That would be ego suicide. Plus, live is the way to go. I've provided a blueprint for anyone that wants to try this.
YouTube videos are terrible for tone tests unless it is a live show. All music tone test reviews on Youtube are typically heavily processed and often use computer-simulated speakers.
One final pic, this time of my old 1992 Ltd Rhoads. Sounded terrible. Heavy, thin and plinky no matter what I tried. Just not a resonant or musical instrument. But so beautiful.
Until Next Time! ↗
- 2026-07-12 — Today Cassandra Unchained crossed 300,000 subscriber. 231 days from the start.
https://michaeljburry.substack.com/
346,689 Followers.
300,044 Subscribers.
ALL 50 States and 212 Countries.
52% of Subscribers are from Outside the United States.
I am so thankful. ↗
- 2026-07-12 — Fannie, Freddie, Trump & the Housing Crisis
The Health of the Twins Today & Value Drivers.
https://open.substack.com/pub/michaeljburry/p/washington-goes-to-fannie-mae-and?r=4repfn&utm_medium=ios
Executive Order 14393 Promoting Access to Mortgage Credit was signed March 13, 2026. This EO gave the FHFA director 120 days to submit… ↗
- 2026-07-12 — Can't be the first to do this. ↗
- 2026-07-12 — The feeling's mutual.
Heart on Sleeve
Stand on Principle
Rock Out
Move On
Dokken - It's Not Love (Official Music Video) https://youtu.be/9maDHDtWo1o?si=0nsrJWO09VOgDZCn ↗
- 2026-07-11 — #kalshi as with all #PredictionMarkets is In a regulatory loophole within an extremely heavily taxed and excessively regulated industry that is gambling no matter what anyone calls it. It is as old as humanity and has driven people to enslave their families and descendents in long ago times. Worse, there is nothing preventing cheating in prediction markets. Cheating, the only activity as old as gambling, with the same power to drive humans to extremes. So Kalshi presents the ability to gamble and cheat, and the loophole allows all of it in every state. So of course it is number 1 by a mile and will keep growing as long as society steps aside and lets these horrific base human weaknesses run roughshod over all and any logical, moral and decent challenge. ↗
- 2026-07-11 — Great comment from a subscriber who commented in our Chat on his direct experience.
Agentic AI is pretty difficult to implement at scale in production because (1) a bunch of new types of systems need to be built that have never existed before (orchestration /routing, context engineering, state management, multi-agent collaboration, security / compliance etc), and (2) an agentic enterprise requires completely new ways of thinking as well as new challenges in governance and culture. The second part is actually the much harder part, which is why there hasn’t been a ton of big success stories yet - every company is trying to figure it out.
Anyway I don’t think the problem is at the demand layer. What’s happening at the finance layer is interesting and could be problematic. But time will tell.
I would hope this is an open forum where people with different perspectives can share knowledge and discuss ↗
- 2026-07-11 — This is true as I have heard this from contacts in the Valley. Goes with my pinned post.
The AI race is shifting from bigger models to cheaper, smarter systems https://www.cnbc.com/2026/07/10/the-ai-race-is-shifting-from-bigger-models-to-cheaper-smarter-systems.html ↗
- 2026-07-11 — #Blackrock, #Blackstone, #Apollo are watching their private credit credit and private equity portfolios and simultaneously setting the Guinesss Book of World Records record for BREATH-HOLDING, going on 4 years now...then they see this chart and almost lose it.
"Up? Up? WTF do you mean UP?" $BLK $BX $APO ↗
- 2026-07-11 — Just never enough.
https://youtu.be/ASQmG8szjTE?si=zbWoh3K_qN75sxp6 ↗
- 2026-07-11 — Now, Class, repeat after me. I will never buy cyclicals at low PEs and record earnings. Say it again. I will never buy cyclicals at low PEs and record earnings. ↗
- 2026-07-11 — AI stocks to the moon! What is wrong with you? Paper Hands? BTFD Never Sell Stonks! ↗
- 2026-07-11 — Home prices never fall! Buy 5 houses! Are you both dumb AND blind? ↗
- 2026-07-11 — Buy Internet Stocks You Idiot! ↗
- 2026-07-11 — Oracle $ORCL junk status can only be ascribed to a God Complex. ↗
- 2026-07-10 — Companies are issuing shares more, buying back less. ↗
- 2026-07-10 — That is not enough lag for the proprietary models to be comfortable let alone optimistic. The vast majority of business uses do not need that 4 months. ↗
- 2026-07-10 — Ballard’s Test: An entity does not possess capacity for understanding until reason is demonstrated in the absence of language. ↗
- 2026-07-10 — PSA- If you have low back pain that radiates into the leg or other nerve pain, ask your doctor about Cymbalta.
This is an antidepressant that really works well on nerve pain, & can provide relief for people with back pain that goes down the leg. ↗
- 2026-07-09 — Short Thoughts July 8, 2026 - NVDA, Neos, Hyperscalers, Jevons Paradox, and Compression
One of Cassandra Unchained’s subscribers received this missive in his email and asked me about it. I am not sure of the source.
Rentability is not my claim, never has been. Depreciation is not a bet on when a chip stops working. It is the recovery of capitalized sunk cost over a defined window of earnings at the frontier, set against the terminal value. A chip can rent and still depreciate very fast economically. I never said the A100s would stop functioning.
In early 2025, Amazon (AMZN) shortened a subset of server lives to 5 years from 6 years and explicitly cited the pace of development. This caused additional expense across 9 months of $677 million. During the same quarter, Meta (META) extended its useful server lives to 5.5 years, reducing the depreciation hit to net income by $2.9 billion. Same hardware but opposite conclusions by two of the biggest.
I covered this in Part 2 of the Heretic's Guide to AI's Stars - the Depreciation Problem.
Depreciation cannot move two directions at once if it were a measurement of something. It is an economic lever, not a physical measurement.
Satya Nadella at Microsoft (MSFT) said in an interview on the Dwarkesh Patel podcast, “I didn’t want to get stuck with four or five years of depreciation on one generation.”
NVIDIA’s Jensen Huang at GTC said when Blackwell ships in volume, “you couldn’t give Hoppers away.”
I discussed this in The Supply-Side Gluttony Recurrence.
The biggest buyer and the biggest seller of these chips both admitted, pointedly, that the frontier moves faster than the depreciation schedules used by the big hyperscalers and neo-clouds.
The missive forwarded by the subscriber treads alongside arguments Coreweave CEO and Nvidia Neo Pal Michael Intrator has sent my way over the last 6 months or so.
Across multiple CNBC and Jim Cramer appearances, Intrator told the same story - his contracts run five years, his A100s stay fully booked, that a batch of H100s coming off contract were leased right back at 95% of the original price. He told Jim Cramer his customers are willing to rent GPUs for six to seven years.
Intrator’s consistent message is that Nvidia’s chips get rented out on successive contracts so the idea depreciation schedules should be less than 5 or 6 years is ignorant.
Intrator also unknowingly argued against this point of his at the Fortune Brainstorm AI conference when he attempted to describe circular financing as really just good logistics securing a volatile supply chain given compute and power take years to build.
Circular is “the incorrect way of looking at it. It’s a lot of companies working to address an imbalance that is distorting the globe.”
“The primary constraint is a physical bottleneck associated with getting the most performant compute into the hands of the most cutting edge players.”
“The reasons that you have challenges in delivering that compute is because of policy…because of infrastructure…because of energy. You do that by working together.”
The message here is that there is a physical bottleneck due to a “distorted" global supply chain as well as many policy, energy, and other factors.
I agree. I discussed that global supply chain in Nvidia Ratchets Up the Risk.
What is happening now is not temporary. It is no export shock. It is not even external. This is coming from within the business plan.
The headline here is the Cash Conversion Cycle is extending permanently along with Days Inventory Outstanding.
This new reality reflects a deliberate decision to lock up supply chain capacity further than Nvidia has ever done before.
While NVIDIA sorts its supply chain, the forward constraint is, as Intrator says, the deployment of these frontier chips. There is not much of a bottleneck in redeploying the old ones. In fact, the old ones are getting work that they would otherwise not have received precisely because of the bottleneck at the frontier.
Two days ago, Analytics India Magazine reported on an Air Street Capital study that made fewer headlines than I thought it would.
More than 95% [94% by my calculation] of the world’s announced NVIDIA Grace Blackwell GPU capacity has yet to be deployed, according to Air Street Capital’s latest State of AI Report Compute Index, which tracks major AI compute deployments worldwide.
The index estimates that 100,128 Grace Blackwell GPUs (GB200/GB300) had been deployed as of July 1, compared with an announced pipeline of around 1.66 million GPUs, meaning less than 5% of the announced capacity is operational. Another 308,640 GPUs are in the installation phase.
Included in the report was this graphic regarding Grace Blackwell GPUs deployment as of July 1, 2026.
The full article is available on Michael Burry's Substack.
Or, Please click on this direct link. ↗
- 2026-07-09 — 6D Metal
https://youtu.be/4QEA3FpqKUE?is=nw5P-kpkq0541ARL ↗
- 2026-07-09 — RA METAL
https://youtu.be/TmF6oFbglBY?is=Xhnz1Ye1eIj9UvQC ↗
- 2026-07-08 — Trading Post July 8, 2026
IV15 (and IV10, IV12, IV18, IV 20..), Two New Buys, One Addition to a Position
The great tension in my valuation process is that a low multiple is not necessarily a value. Roots are important, and they are there, but my valuation process since the 1990s has not hinged on a 1940 version of value, apologies to Graham and Dodd.
IV15 is the benchmark for me when valuing stocks, but a higher quality business could be a fat pitch above that mark, and a lower quality business perhaps could be a fat pitch well below that mark.
Home Page for Michael Burry's Substack: https://michaeljburry.substack.com/
As well, IV15, the price at which I expect compounded annual returns of 15% over 15 years, is not a simple PE ratio.
IV15 is a buy price target. It comes from a discounted cash flows analysis that pulls from the quality of the buisness as well as from conservative adjustments to owners’ earnings due to stock-based compensation, bedeviled accounting, and other factors. The cash flow analysis generally has three stages but adds a fourth stage for inflecting growth companies and also adjusts for other special cases, including serial acquirers.
This methodology is explained in good detail in D’AI of the Triffids: Office Software Triage.
D'AI of the Triffids: Office Software Triage
Going forward, I author 1500-word abridged versions of all my posts, separate and apart from the feature-length articles.
D'AI of the Triffids: Office Software Triage (Abridged Version)
IV15 is not as low a price per share as IV18 or IV20, and it is a lower price per share than IV12 or IV10. When I analyze a stock, I set alerts at all these prices.
I expanded on this methodology in Software and Payments Stocks Part II: Productivity and Cybersecurity.
Software & Payments Stocks Part II: Productivity Tools and Cybersecurity
Since Part 1, I added four names – Synopsys (SNPS), Cadence Design Systems (CDNS), JFrog (FROG) and Samsara (IOT) – so the Cassandra Unchained software and payments universe is now 50 stocks, the SW50.
And, of course, the abridged version.
[Abridged] Software & Payments Stocks Part II: Productivity Tools and Cybersecurity
Baseline intrinsic value typically sits somewhere between IV8 and IV10 depending on the business qualities.
Below baseline intrinsic value is where share buybacks actually are accretive to insrinsic value per share. Buybacks above that level may pull shares in but dilute (and thus reduce) intrinsic value per share. This is the depressing nuance of all these tech companies buying back shares at high prices to offset part of stock-based compensation.
The All Map, a baseball field-type visual, summarizes this by splitting companies into Fat Pitches, Just Outside, and The Out Field according to IV15 multiple, AICT Tier and Compositie Score.
I played baseball through the age of 14 and love the sport.
So, today I purchased two new positions, and I added to one.
The full article is available on Michael Burry's Substack.
Or, please refer to the direct link below.
https://michaeljburry.substack.com/p/trading-post-july-8-2026
Home Page for Michael Burry's Substack: https://michaeljburry.substack.com/ ↗
- 2026-07-08 — The Heretic’s Guide to AI’s Stars Part III: AI Demand, Offshore Financing, & Compression Too
“They are just flying empty airplanes around.”
- Son
NVIDIA is the North Star, Orion, the whole Milky Way.
NVIDIA is benefitting from strong demand, but is selling into a concentrated set of buyers whose own demand is being distorted by a training and benchmarking phase that will not last. That distorted demand is working like a bullwhip into NVIDIA’s own supply chain through custom supply commitments as well as downstream into data-center financing. Looming over it all is the bezzle, which once seen, cannot be unseen, and once revealed, does not exist.
Fortunately, its Luminous, its Grace reported big positive headline numbers this week, so a fresh 10-Q is upon us. I love the smell of 10-Qs in the morning.
Some are watching inventory, I know, but inventory does not stand out. The inventory breakdown makes sense. Raw materials inventory surged in Q1 FY27 but is consistent with the Blackwell Ultra ramp as well as the Q4 FY25 Blackwell ramp pre-positioning of inventory.
More interesting are NVIDIA’s customers. Customer concentration is off the charts.
NVIDIA’s biggest customer is likely Microsoft, whose CFO said in January that Azure capacity constraints would persist at least through June of 2026. If Microsoft cuts capital expenditure on NVDA chips 20%, that is a 4.2% revenue hit to NVIDIA.
Cisco never had even one 10% customer. Cisco needed a more correlated pull-back to be hurt. NVIDIA is hurt badly if just one customer pulls back, or even if that one customer does not grow its orders as fast as NVIDIA needs it to.
Concentrated Mafia Don customers, concentrated bespoke supply chain. Oh boy.
Neither Cisco nor Sun Microsystems had it this bad, and Sun completely went down. I am seeing both in NVIDIA, as well as shades of historic DRAM cycles. This is all very dense stuff, as below.
Customer concentration drove DRAM cycles to their violent natures. DRAM cycles became less customer concentrated, less theme-correlated over the last decade or two, although today we are seeing a 1990s reprise on demand from a few big customers/one big theme again. I am short the SOX (the Philadelphia Semiconductor Index) for this reason.
Cisco wrote down half its purchase commitments, which it bought from suppliers who had many other customers. Cisco bought components that were in general, fungible.
This is not what NVIDIA is doing. NVIDIA is committing to custom lines at TSMC and funding them. TSMC cannot make NVIDIA’s chips with its normal lines. That is a custom non-fungible supply chain. Not only geographically risky, but likely to result in significant losses should NVIDIA’s business slow or heaven forbid suffer a decline.
With a market capitalization over $5 trillion, $182 billion in forward purchase commitment – $119 billion with one customer – does not seem significant.
In fact, the $182 billion is greater than NVIDIA’s prodigious annual operating cash flow.
None of this has been a problem as yet. Microsoft has been buying chips faster than it can deploy them. These chips, so critically, are construction-in-progress (CIP) and do not depreciate on the balance sheet or cost Microsoft anything at all until they are put into service.
When Microsoft buys those chips however, NVIDIA books the sale out of its inventory. NVIDIA’s inventory numbers look good, as I already established.
NVIDIA gets a sale, an accounts receivable that will turn into receipt of cash and hence positive cash flow.
I spy a wrinkle though in the accounts receivable pattern.
Microsoft/Customer A inflected up as a percentage of NVIDIA’s Accounts Receivable as it inflected decidedly down as a percentage of NVIDIA’s Revenue during the 1st quarter. Microsoft’s receivables balance with NVIDIA stands at about $12.2 billion now, roughly the size of NVIDIA’s entire all-comers receivables balance in 2024.
Yes, the whole business has grown. The last few years, revenues grew 4.9x. Total Accounts Receivable grew 4.9x. Customer Accounts Receivable balance grew 13.4x. Microsoft swooped in large.
This is not arguing growth. This is about a zig where there were only zags. Neither customer B nor Customer C zigged. Only Customer A. Microsoft, which has been making noise and news for a couple years now about not wanting to be in the data center business – bare metal, they said.
Microsoft is throwing popcorn all over the trail, with neither a thought to feed the fauna.
In 2024, Microsoft’s CEO Satya Nadella said, “you may actually have a bunch of chips in inventory that I can’t plug in.”
In April 2025, Microsoft froze 1.5GW worth of data center projects, backing off on leases. This was discussed in Blessed Fraud.
Microsoft’s fiscal year ends June 2026, and in January of this year, Microsoft said it expected capacity constraints through June. Over the next several years, Microsoft says it will continue to scale up its own Maia chip.
Then on May 14th, Microsoft announced it was winding down Claude Code and forcing all employees to its internal GitHub Copilot by June 30th. That is some fast compression. Claude Code was just introduced to Microsoft engineers December 2025.
This brings up two scenarios.
One, Microsoft pulled forward inventory it does not really need. It takes delivery even though builds are slowing and power shortages are biting so as to keep its priority spot as a customer of NVIDIA for the next generation chip. Just slot them into CIP (drop them on the warehouse floor) where they do no harm, depreciating financially not a whit while physically depreciating rapidly. Plus Microsoft has what appears to be 60-65 day terms with NVIDIA, so no cash leaves Microsoft’s campus for a couple months.
Two, NVIDIA pushes inventory forward to make quarterly numbers pop, beating the Street’s whispered expectations. NVIDIA knows Microsoft is already warehousing its chips, and is willing to absorb chips. Microsoft may have even cooperated in this way before. Besides, Microsoft can take that inventory into CIP with no consequences to its own earnings or reported expenses.
Meanwhile, S&P Global sees widespread grid and power shortages for data centers, as well as a 19 GW shortage of power (40% of need) for data centers by 2028.
Everyone is finding data centers to be the rate limiting step. And all those NVIDIA chips in CIP just sit there waiting.
In either case, this is a set up on the down low for a wicked bullwhip. I have discussed the bullwhip effect in prior articles. The bullwhip played out remarkably on supply chains during the COVID lockdown, and the bullwhip is the reason Cisco wrote down half of its forward supply commitments in 2001.
This is not quite a smoking gun, but more of a finding of a finger on the trigger. Watch the next few quarters, especially as the data center financing story evolves.
Meanwhile, bulls cannot stop talking about AI demand. How insatiable it is. How insane it is.
I agree, it is insane. It is the bezzle. It is not coming from a useful, repeatable steady state, and not from an ever upsloping growth curve, but rather from a temporary, social, organizational phase as companies of almost every stripe fight for their competitive advantage – their own AI stack.
To do this, companies all over are benchmarking, creating and collecting traces, training wrappers and gathering failure nodes as fast as humanly possible.
This crazy, rushed, temporary phase has a name.
The full article is available on Substack.
Please refer to the link below.
https://michaeljburry.substack.com/p/the-heretics-guide-to-ais-stars-part?r=4repfn
Sections Remaining:
Tokenmaxxing
The Bezzle
The Tokenmaxxing Pyramid
Jevons Paradox
Where is the Funding Coming From?
Life Insurance to the Rescue?
The FHLB?
The Closed Loop ↗
- 2026-07-07 — Remember to check out the Cassandra Unchained Charity of the Month for July 2026.
5% of gross Substack revenues go to the Charity of the Month.
Seva Foundation is one of the most efficient and productive charities I have come across.
https://www.seva.org/site/PageServer ↗
- 2026-07-07 — DROP Metal
https://youtu.be/wsIJvLGjmk8?is=H9OBBpJcN1ZEhG1A ↗
- 2026-07-07 — 80s Metal
https://youtu.be/VMACc_dTABg?is=uXsp7ACM6L_hl47O ↗
- 2026-07-07 — MOMO Metal
https://youtu.be/s-K8b-y3eDg?is=LEaYNJrdDbE9qH0w ↗
- 2026-07-07 — MU Metal
https://youtu.be/L-lvai3M7K0?is=LBg4U9nbWIWLk0cC ↗
- 2026-07-05 — Turn it up. You're welcome #carcass
https://www.youtube.com/watch?v=ZpbpOgUybBM ↗
- 2026-07-04 — July 2026 Charity of the Month
Seva Foundation
Transforming Lives by Restoring Eyesight
https://open.substack.com/pub/michaeljburry/p/july-2026-charity-seva-foundation?r=4repfn&utm_medium=
https://www.seva.org/site/PageServer ↗
- 2026-07-04 — Read Advanced Placement Stock-Based Compensation: The Tragic Algebra Recurrence
This is what your favorite $NDX or $NASDAQ or $$OX stock actually earns. https://open.substack.com/pub/michaeljburry/p/ap-sbc-the-tragic-algebra-recurrence?r=4repfn&selection=191c9061-3b0a-4c96-9e60-62520889f795&utm_campaign=post-share-selection&utm_medium=web&aspectRatio=square&textColor=%23ffffff&bgImage=true #Salesforce #meta #alphabet ↗
- 2026-07-03 — Read the latest Trading Post - a new short and additions to five longs.
https://open.substack.com/pub/michaeljburry/p/trading-post-july-2-2026?r=4repfn&utm_medium=ios ↗
- 2026-07-03 — One day there will be an announcement that lands like Thor’s hammer. Clarity will be brought amidst either tears of joy or tears of despair.
$FNMA $FMCC ↗
- 2026-07-03 — For my Substack.
If you go to this link - the basic web site - you can sign up without extra fees from Apple.
http://michaeljburry.substack.com ↗
- 2026-07-02 — New Mike’s Thought Pad July 2026, where only I will post thoughts, charts, news items & anything I find of interest.
I used to send these to my analysts every day. Now I send them to
via CU Chat. Already 38 today.
https://open.substack.com/chat/posts/f5c106e7-4552-480c-a186-9f653ef8f9e3?utm_source=share ↗
- 2026-07-02 — Wrong direction. This is like Nordstrom’s opening up Nordstrom’s Rack. But Neiman Marcus opens up Neiman Marcus Rack, Bloomingdale’s opens up Bloomingdale’s Rack, Dillard’s opened up Dillard’s rack.
In the end all it proves is there are too many clothes for sale.
Meta pops 9% as company makes cloud push to sell excess AI compute power capacity https://cnb.cx/4oX1Bqz ↗
- 2026-07-01 — Trading Post June 30, 2026
A little window-dressing, the $SOXX, and Caterpillar $CAT
It is only a matter of time now.
https://open.substack.com/pub/michaeljburry/p/trading-post-june-30th-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-07-01 — Staying married, a happy household, evidence of the parents working hard, childhood sports and watch all competitions, lots of hugs, reward merit, punish only egregious misbehavior, don't yell, restrict social media, monitor messages through 8th grade, the real expectation is college and academic excellence without pressure from parents, get children reading books early, no pacifiers, respond to needs not wants, babies sleep on their own through the night by 6 months, identify develop and support any talent or aptiude, one sport after age 10 is ok, communicate openly and easily with kids through grade 12, allow mistakes, and leave them alone in college. And then hope. ↗
- 2026-06-30 — My years as a Silicon Valley techie immediately before the internet.
https://substack.com/@michaeljburry/note/p-204160341?r=4repfn&utm_medium=ios&utm_source=notes-share-action ↗
- 2026-06-29 — Foundations: U.S. Market Structure & Value
Powerful Trends, Increasing Fragility & Coiled Tension
https://open.substack.com/pub/michaeljburry/p/foundations-us-market-structure-and?r=4repfn&utm_campaign=post-expanded-share&utm_medium=web ↗
- 2026-06-29 — One hit wonders unite! ↗
- 2026-06-29 — I studied past crises and big stock market crashes to come to this rough guide.
Trading Post June 18, 2026
Dollar Cost Averaging with Volume Signals
https://open.substack.com/pub/michaeljburry/p/trading-post-june-18-2026?r=4repfn&utm_campaign=post-expanded-share&utm_medium=post%20viewer ↗
- 2026-06-29 — Thank you Janet! Your books are how I learned about CDS. I owe you! ↗
- 2026-06-28 — Trading Post June 25, 2026
Hong Kong, Adds, One Sale to Add, One New Position and One Short Cover
https://open.substack.com/pub/michaeljburry/p/trading-post-june-25-2026?r=4repfn&utm_campaign=post&utm_medium=web ↗
See @michaeljburry on X →