원문 보기
Michael Burry just did it again. The man who saw 2008 coming when nobody else did, the guy Christian Bale played in The Big Short, just spent weeks reading 1,017 audited annual reports. And what he found is a $1.7 trillion lie sitting at the heart of the stock market. He is now naming nine specific companies that he says have transferred more wealth to their employees than they have ever earned for shareholders.
Nine names that millions of people own through their retirement accounts and don't even know it. If you own any one of these stocks or any index fund that includes them, you need to see this list. Because if Burry is right, the math behind your portfolio is not what you think it is. Before we get to the nine names, you need to know why this particular guy saying this particular thing is a big deal.
. Between 2000 and 2008, his original fund returned 489.34% net of fees, almost entirely off a single contrarian unfashionable bet against the US housing market. Wall Street thought he was crazy. He was right. happened. Then, in November 2025, he did something nobody saw coming. He shut down Scion entirely, deregistered the fund with the SEC, .
Today, that Substack has over 268,000 subscribers and ranks as the second largest finance newsletter on on platform. Here is why that matters. For his entire career, Burry was . Now, he i April 7th, 2026, he posted what may be the most consequential piece of forensic accounting any retail investor has ever seen for free.
He calls it the tragic algebra recurrence, a mental model he says is three decades in the making. And inside it, hidden in plain sight, are the names of nine companies he believes are quietly mortgaging the future of every shareholder who owns them. Here is the core idea, and it is genuinely shocking once you see it clearly.
When a tech company pays an employee in stock instead of cash, what's called stock-based compensation or SBC, Wall Street treats that expense as essentially free. Analysts add it back to earnings. Companies highlight adjusted non-GAAP profit numbers that exclude it. The justification you've heard a thousand times is that it's non-cash, so it shouldn't count.
But Burry's whole point is that it absolutely is cash, just delayed. When those shares vest, the company has to do one of two things. Either it lets the share count balloon, which dilutes every existing shareholder, or it goes into the open market and buys back the same number of shares to offset the dilution, which is real cash leaving the business right now.
Either way, you, the shareholder, lose. You either own a smaller slice of the same pie, or the pie itself is smaller because the company spent its money plugging the hole. Burry built a formula that captures all of this. He took 10 years of audited 10-K filings, every cash flow statement, every footnote, every shareholder equity table across 97 of the 100 companies in the Nasdaq 100 index.
He calculated the true cash cost of SBC for every single year. And the numbers he came back with are, to put it mildly, hard to swallow. Of every dollar of GAAP earnings on the Nasdaq 100, shareholders only see 83.49 cents. The rest, that missing 16.51 cents, doesn't disappear. It just travels quietly from your pocket into employees' pockets through the share count.
And on Wall Street's preferred adjusted earnings, the numbers CNBC quotes, the numbers your brokerage app shows you, Burry calculates the gap is even worse. Wall Street is overstating the true earnings of the entire Nasdaq 100 by 42%. That is not a rounding error. That is the difference between thinking a stock is fairly priced and realizing it is in a bubble.
Now, here is where it gets specific. Within those 97 companies, Burry identified a cohort he calls the tragic tier. These are companies where, over the entire 10-year measurement period, the cumulative true cost of stock-based compensation actually exceeded their cumulative earnings. Owners' earnings for all of them, in aggregate, are negative.
As Burry put it, "These nine have proven over the time they have been public on this green earth that they can issue stock to a degree that the true cost of stock-based compensation can be expected to exceed earnings." The nine names are Tesla, Palantir, CrowdStrike, DataDog, Axon Enterprise, Workday, Zscaler, Shopify, and Marvell Technology.
Let's go through them. Number one is Tesla. Burry calls Tesla the lord of the tragic tier, and the reason is so extreme it warps the entire data set. Removing Tesla alone drops the overall Nasdaq 100 overstatement from 20% to 12.5% meaning a single stock is responsible for nearly half of the entire index's earnings illusion.
The 2018 Musk pay package was already $56 billion at grant. The newly approved 2025 package, if all milestones hit, could be worth roughly $1 trillion. Burry Burry says even Tesla's own epic dead weight pales next to what is coming. Number two is Palantir. This is the one that should worry the retail crowd most because Palantir is one of the most owned stocks among individual investors today.
The bull case is government contracts, AI software, and Karp's vision. The hidden cost is that Palantir issues an enormous portion of its revenue back to employees as equity. The non-GAAP profitability that gets quoted in the press strips that expense out entirely. Once Burry's adjustment is applied, the underlying owner's earnings shrink dramatically.
Number three is CrowdStrike. Cybersecurity is famous for using equity to keep top engineers from defecting to competitors. CrowdStrike is the extreme version of that strategy. The non-GAAP earnings that fund managers price the stock on are propped up almost entirely by the SBC ad back. Strip that back out, factor in the cash spent absorbing dilution, and the actual earnings power available to a shareholder of CrowdStrike is a fraction of what the screen says it is.
Number four is DataDog. DataDog is the textbook example of the GAAP versus non-GAAP GAAP Burry is attacking. The company can post a GAAP operating loss in a given year, then turn around and report a giant adjusted profit by adding back hundreds of millions in SBC plus the payroll taxes those equity grants generate.
The defense DataDog gives in its filings is that SBC is unrelated to operating performance. Burry's response is that compensation is the most fundamental operating expense any business has, period. Number five is Axon Enterprise. Axon makes tasers and police body cameras. Operationally, the business is excellent and growing fast.
But Axon also runs one of the most aggressive equity compensation plans in the entire index with a CEO performance award structure tied to volatile stock price milestones. The result is that Axon's GAAP earnings whip around violently from year to year based on how those grants are recognized. And the non-GAAP clean numbers investors actually look at hide the true labor cost almost completely.
Number six is Workday. Workday's interesting because the company itself basically admits the problem. In their own financial outlook releases, Workday says it cannot reliably forecast its GAAP earnings because the volatility of stock-based compensation and related taxes is too unpredictable. Think about that for a second.
The company is openly telling you that the actual cost of paying its workforce is so large and so erratic that even Workday can't model it. And yet, Wall Street prices the stock as if the non-GAAP post-adjustment numbers are reality. Number seven is Zscaler. Same playbook as DataDog and CrowdStrike.
Zscaler explicitly defines its non-GAAP operating income by stripping out stock-based compensation and the related payroll taxes. Hundreds of millions of dollars of real labor cost get swept off the income statement and onto the share count where retail investors don't see it. The valuation premium investors pay for Zscaler is built on numbers that pretend that expense doesn't exist.
Number eight is Shopify. Shopify is the most insidious one to spot because the company recently announced a $2 billion share buyback program that on the surface looks like it's returning capital to shareholders. It is not. Read Burry's analysis carefully and you'll see what's actually happening. The buyback is sized almost exactly to absorb the new shares being issued to employees through SBC.
The cash leaves the business. The share count stays roughly flat. Shareholders get nothing. Employees get the equity. The headline says buyback. The reality is dilution offset. And number nine is Marvell Technology. Marvell is the lone semiconductor name on the tragic tier list and it is also a name caught in the crossfire of the broader Nvidia driven recalibration.
According to Amelius research analysis, if Marvell were forced to include SBC in its non-GAAP earnings the way Nvidia just announced it will, the hit to Marvell's reported EPS would be between 14 and 20%. That is the size of the illusion sitting inside Marvell's quoted numbers right now. Put those nine companies together and the math gets uncomfortable fast.
Over the 10-year window Burry studied, the nine collectively reported $42 billion in GAAP net income. But their true owners earnings, adjusted for the real cash cost of SBC, came to negative $267 billion. Read that sentence again. Negative $267 billion. As a group, these nine companies have transferred more wealth out the back door to their employees than they have ever generated in actual cash profit for the people who own the stock.
Now, zoom out from those nine to the whole Nasdaq 100. The numbers Burry pulled together are almost cinematic in scale. Over the decade ending in fiscal 2025, the 97 primary constituents of the Nasdaq 100 reported 4.9 trillion dollars in cumulative gap net income. Wall Street's adjusted version of that same earnings number, after analysts add back SBC and other items, came to 5.8 trillion dollars.
Burry's true owners' earnings, calculated through the tragic algebra, came to 4.1 trillion dollars. That gap, 5.8 trillion dollars versus 4.1 trillion dollars, is the famous 1.7 trillion dollar earnings illusion. And that is the single number you should remember from this entire video. 1.7 trillion dollars of earnings that Wall Street has been telling investors existed that simply, mathematically, did not.
It's not opinion, it's not a forecast, it's the difference between what was reported and what was real, summed across an index, audited to the dollar by 97 different accounting firms. In practical terms, here is what that means for the price you pay to own Nasdaq stocks today. When CNBC tells you the index is trading at, say, 30 times forward earnings, sounds expensive, but tolerable, right?
Apply Burry's correction. That same 30 times on Wall Street's adjusted number becomes 42.6 times on real owners' earnings, not approximately, exactly 42.6 times. 30 multiplied by the ratio of the inflated earnings to the real ones, the market isn't expensive. The market is, by this measure, in a 1999-style valuation regime, and almost no one is talking about it.
Even individual names you would assume are clean show up dirty under this lens. Lamb Research overstates its profits by 49.1% by Burry's calculation, meaning shareholders capture only 67 cents of every reported dollar. Nvidia itself overstates by about 28.9% driven by what Burry calls a buyback to nowhere, share repurchases that look like capital returns, but are really just absorbing the dilution from RSUs vesting in the background.
AMD and Intuit clock in at 32.6% Synopsys at 30.8% Netflix at 22.9% Amazon at 21% The illusion is everywhere. Here is the part that turns this from an academic accounting argument into an actual market event. On February 25th, 2026, deep inside Nvidia's Q4 earnings release, Nvidia's CFO Colette Kress announced that beginning in the first quarter of fiscal 2027, Nvidia will include stock-based compensation expense in non-GAAP financial measures.
One sentence buried in the commentary, and it changed everything. Nvidia could afford to do this because Nvidia is a printing press. The cost of stock-based comp at Nvidia rose from $4.7 billion in fiscal 2025 to $6.4 billion in fiscal 2026. But against $116 billion in after-tax profits, the hit to non-GAAP EPS is only about 3%.
Tiny. Nvidia could absorb the honesty. Most of the rest of tech can't. According to that Melius Research note again, applying the same accounting standard would mean a 14 to 20% hit to non-GAAP EPS at Broadcom, AMD, and Marvell. That is not a small disclosure change. That is a quiet, structural earnings reset hitting tech in slow motion.
And if you want the historical sanity check, Warren Buffett has been screaming about this for 30 years. In his 2018 chairman's letter, Buffett wrote that managements who claim stock-based compensation should not count as an expense are essentially asking, in his words, "What else could it be? A gift from shareholders?" Burry's article is, in many ways, just Buffett's ancient complaint finally backed up with a thousand pages of audited data.
Here is the math that makes this dangerous, not just annoying. When a shareholder receives only 83 cents of every dollar earned, the missing 17 cents doesn't sit there. It compounds against you every single year. Burry actually ran the numbers on what that means over time. At the NDX 97's measured delta E of 83.49%, real intrinsic value per share grows at only 58% of GAAP's reported pace over 5 years.
Only 16.5% of it over 10 years, and just 2.7% of it over 20 years. Read that one more time. After 20 years, an investor in the Nasdaq 100 captures, by Burry's math, only about 2.7% of the GAAP intrinsic value the company appears to be creating. The other 97% is, has been, and will continue to be diverted to employees through equity grants.
This is not a recession risk. This is not a fed cuts or doesn't cut question. This is structural. It runs continuously in good markets and bad, and it gets worse as the businesses get larger because the SBC is a percentage of revenue. And the cruelest part of all is that you cannot escape it by buying an index fund.
If you own QQQ, you own Tesla. You own Palantir. You own all nine names on the tragic tier with full index weight. The structure of the index doesn't let you opt out. You buy the whole basket, including the cesspool, as Burry bluntly put it. The only way out is to look at the names individually and decide which ones you actually want to own based on what they really earn, not what they say they earn.
So, pull all of this back together. One of the most successful contrarian investors of the last 25 years, the man who saw the housing collapse coming when literally no one else did, just spent weeks in 17 audited filings and came back with a list. Nine companies, he says, shareholders have been quietly subsidizing for a decade.
Tesla, Palantir, CrowdStrike, Datadog, Axon, Workday, Zscaler, Shopify, and Marvell. A $1.7 trillion gap between what was reported and what was real. A Nasdaq that, on his math, is trading closer to 42 times earnings than 30. And a structural problem that compounds against shareholders every year it goes unaddressed.
Burry is not always right. His timing has been early before. His 2023 short against the S&P and Nasdaq closed at a loss. That is a real, documented risk, and you should weigh it. But, the framework he's using here, audited numbers, 10 years of data, the same forensic style that let him see 2008 is not speculation.
It's arithmetic. And the arithmetic says that the non-GAAP earnings the entire industry has been reporting for the last decade are, in Burry's word, an illusion. Whether that illusion ends in a slow grind or a sharp crack, nobody knows. But once you can see it, you can't unsee it. And you can decide, name by name, what you actually want to own.
Comments 0