Michael Burry, a hedge fund analyst and subject of the film, “The Big Short” who became famous for “shorting” Mortgage Backed Securities, exotic financial instruments responsible for the 2008 financial crisis, made headlines this week when he accused Big Tech companies of overstating their AI earnings through creative accounting and “fudging depreciation schedules”. Burry then put his money where his mouth is and made a $9,200,000* bet these securities would inevitably crash. He also deregistered his hedge fund, presumably to avoid the headache of the inevitable blowback from his investors and critics. There’s no question Burry is a savant when it comes to investing, but I wanted to write a post on why I think he got this one wrong.
Depreciation Schedules
Under GAAP standards, when you report your operating expenses every quarter as a public company or GAAP filer, you take your CapEx, anything that you make a big investment in, and depreciate it over some lifetime. There’s an accounting process by which you can do an internal review and determine that, if your depreciation schedule doesn’t actually map to reality, you should make some sort of adjustment and there’s a periodic assessment that’s done to do just that. For example, in the last 12 months, Google, made $120 billion in operating profit.
So how does this periodic assessment work? Let’s assume Google makes $70 billion in CapEx this year. That CapEx does not get deducted from their operating profit when they report earnings. You take that $70 billion and divide it by what’s called its useful life, which you then report every year. If you do this over 3 years, it’s about $24 billion a year versus 6 years at $12 billion a year. This would impact the operating profit by, call it 10%, if the difference is between 3 and 6 years.
What Michael Burry is saying is the hyperscalers have extended the depreciation schedule, or the useful life of their data centers, by roughly 2x, which cuts the operating costs in half when they report it and, in turn, inflates their earnings. Using this logic, Burry’s claim is that they’re cooking the books. That’s all well and good, except for the fact that Google addressed this change in Q1 of 2021, where they said their servers are now going from 3 to 4 years.
The chart referenced above from Michel Burry’s twitter, shows the combination of servers and networking equipment. Separately, in 2021, Google took networking equipment from 3 to 5 years and then in 2023, from 5 to 6 years. This is a result of the effort where they did an analysis, so why the sudden change? What happened?
What happened is that the data centers transitioned from being primarily data storage and data transfer systems, where you would use hard drives, RAM, and memory to store data and then transmit it back out, to being data processing centers because of the AI boom. As AI became more important in the data center, more of the dollars were allocated towards three different components: Chips from data storage, which were initially hard drives, which you throw away every 18 or 36 months, solid state memory, which would last 2 years, 4 years, or 6 years, depending on the type you were using, and finally RAM, which had this kind of variable lifetime as well.
Suddenly, when you put these processors in to process the data to do AI, the majority of the spend and energy was going towards the processors. Google’s head of AI and Infrastructure recently discussed where they were in the CaPex spend cycle at a conference.
“We’re early in the cycle, is what I would say, certainly relative to the demand that we’re seeing. So our internal users are, we’ve been building with TPUs for 10 years, so we have now seven generations in production for internal and external use. Our seven and eight-year-old TPUs have 100 percent utilization.”
This is an important technical nuance. These 7 and 8 TPUs and GPUs that are sitting in the data centers are still being used, AND they’re being used at 100 percent utilization. So this actually justifies and validates the depreciation schedule being much longer versus shorter.
On top of this, the business models of these companies are far too good for them to get to the point of having exhausted every other operational tactic that they then have to cook the books. If it were one shady outlier, maybe… but these are not the 7 companies, at that point an industry, that are going to commit financial fraud.
The practical thing that’s happening that Burry doesn’t understand is that there are multiple meaningful iterations on the underlying tech, such as how kernels are working, how the attention mechanisms of these models are being rewritten, how technicians are swapping out HBM for SRAM, in these designs, in how they’re building, in some cases, really huge dyes, in some cases, much smaller chiplets, and so on and so forth… All of this creates more and more utilization. And in turn, these servers last longer and need more compute. I appreciate that the circular nature of these AI deals has a lot of people spooked, but to accuse an entire industry of financial fraud seems glib and overreaching-especially when these products are creating real, measurable value. History doesn’t repeat, but it does rhyme… But I just don’t think you apples-to-apples comparison to the financial engineering of the 2008 crash.
I think that in order to make these accusations, you need to have some modicum of technical grounding that I just don’t believe Burry has (and that’s not a criticism. It almost makes no sense for Michael to become technically literate when the world of short selling demands its own specialized set of skills and there are only so many hours in a day).
But what about the sudden wave of investors shorting Big Tech?
Here’s the thing with shorts in general, and I don’t like shorting, so I’m admittedly biased: They’re supposed to be a check on financial malfeasance. When you look at these short selling firms, for every one of them that actually uncovers malfeasance, what it really is, is the creation of chaos and innuendo under the guise of their self-asserted right to free speech. What they do is put out some screed that tries to move the market and, thusly, positions them against a stock before it even comes out. At this point, they hope to close out their position and make some money. Frankly, I think that’s a pretty sad and terrible way to live one’s life… but it’s legal and so they’re allowed to do it. Another important point, is that shorting takes a TON of capital and an equal amount of patience. The fear, uncertainty, and doubt you can create if you’re Michael Burry and you shorted the housing crisis correctly, took TWO years for him to validate. In defense of the short sellers, these things are scary and painful to execute.
In short (see what I did there), my point is shorting is just somebody’s ability to cry fire in a theater, and, quite honestly, it’s extremely hard, if not impossible, to commit financial fraud as a public company in 2025. Obviously, you have your Enrons and probably some one off examples from 2008, but here we are, at the end of 2025, and the past is not prologue.
Another nuance inherent in shorting is the assumption that a company is worth their historical sales numbers. I don’t think that’s how shareholders often do, or perhaps should, think about what they’re buying, which is an ownership interest in the future of the enterprise. When you invest in a nascent. business, you’re not saying, “hey, that thing is worth what the employees did last year before they even started the company.” You’re making a bet on the future potential of the business and what you think the cash generation over time will be. Your time horizon may be different than mine, and that’s how a market finds a price. As a result, I think there’s probably a market trying to find a price for a company like Palantir or even Tesla, where people have a great deal of difference in opinion over what the future potential of the business is and, as a result, what the earnings generation will be at different time scales in the future. That’s how they’re getting to the current market price, so who am I to judge? I’m just the person who would make my own decision of my own time scale and my own estimation of the future of that business if I were putting my own capital into said business. If I haven’t studied the business well, I really can’t justify my point of view or opinion on the value of the business relative to its future earnings potential.
This should probably be a separate post altogether, but that’s how I would look at it. I would make an investment for the long term if I were buying the shares. Not look at the last year’s numbers and say that there’s a valuation arbitrage opportunity and that’s what I’m buying. But whatever. To each their own. Over time, the market corrects itself and, theoretically, that’s the value of the free market and how capitalism is supposed to operate.
That all being said, I think Michael Burry’s got this one wrong.
*a funny aside: Media outlets reported Burry’s short at just under a billion dollars. In options trading, you get 100 shares per option, so Burry only spent 1 billion the media reported. Burry called out their financial illiteracy in this tweet.