UpTrajectory Review

Michael Burry, the investor who built his reputation shorting the housing bubble, has turned his attention to the AI buildout and put a date on the reckoning: 2028, possibly stretching into 2029. In a Substack post published Thursday, September 24, 2026, Burry argues that Big Tech has signed the corporate equivalent of a 30-year mortgage on AI infrastructure, and the terrifying part is not the monthly payment but the roof that starts leaking while the loan still has decades to run. His core claim is that by 2028 or 2029, the accumulated AI commitments at companies like Microsoft, Alphabet, and especially Oracle will be so enormous that even a relatively small write-off will land with a force the market cannot currently price in.

Burry's framework is the capital cycle, and the numbers he cites are stark. Net capital investment by S&P 500 companies, meaning capital spending minus depreciation, hit roughly 2.07% of GDP as of June 30, 2026. In nearly four decades of data, only the aftermath of the March 2000 Nasdaq peak exceeded that level. After the dot-com buildout collapsed, depreciation and write-downs kept S&P 500 net investment negative for twelve consecutive quarters, from mid-2003 through mid-2006. Burry's point is that write-offs lag the spending peak by a few years, so the pain arrives well after the headlines about record capex have faded. He expects spending to keep climbing for a few more quarters before the cycle turns.

What makes this more than a generic bubble call is the specificity of Burry's accounting critique. At Microsoft, he counts uncommenced leases that nearly tripled to over $300 billion, a figure he connects to the company's decision to stretch data center useful lives to 25 years, both of which have the effect of shrinking reported capital spending on the income statement. At Amazon, leases and purchase commitments rose 81% in nine months to roughly $267 billion. Burry's argument is that the true scale of AI spending is being obscured by lease accounting and depreciation assumptions, meaning the headline capex numbers understate the real commitment. Among the world's most profitable companies, he exempts only Apple, writing that the rest are betting everything on AI as their free cash flow turns negative and their borrowing accelerates.

For a small-business operator, the temptation is to file this under Wall Street drama and move on. That would be a mistake. If Burry's timeline holds, the downstream effects reach far beyond tech shareholders. A 2028 write-off cycle at the hyperscalers would likely trigger a broad repricing of AI-exposed assets, tighter credit conditions for anything touching data center construction, and a wave of layoffs at the contractors, chip suppliers, and commercial real estate players who have built their own expansion plans around Big Tech's spending. Oracle, which Burry places squarely in the crosshairs, has already cut more than 2,500 workers even as its AI spending surges, a pattern that could repeat across the ecosystem if the capital cycle turns.

We are skeptical of precise date predictions, and Burry himself has a mixed record on timing, most famously covering his subprime short too early. The capital cycle logic, however, is harder to dismiss. Every infrastructure boom in modern economic history, from railroads to fiber optics, followed the same arc: spending peaks, depreciation catches up, write-offs follow, and the companies that looked invincible suddenly look overleveraged. The counterargument is that AI revenue is real and growing, unlike the eyeball metrics of 1999, and that companies like Microsoft and Alphabet generate enough cash to absorb losses that would bankrupt lesser firms. That is true, but Burry's point about negative free cash flow and accelerating borrowing suggests even the giants are stretching.

What to watch: quarterly capex reports from Microsoft, Alphabet, Amazon, and Oracle over the next four to six quarters, particularly whether depreciation schedules keep lengthening and whether lease commitments continue outpacing revenue growth. If you run a business that depends on AI vendor pricing, cloud contracts, or tech-sector employment in your local economy, treat 2028 not as a prophecy but as a planning scenario. Lock in favorable contract terms now, avoid building your cost structure around the assumption that AI compute prices keep falling, and pay attention to whether the hyperscalers start quietly revising their useful-life assumptions downward. That is usually the first crack in the wall.

“The scary part of a 30-year mortgage is rarely the payment. It is the roof that wears out while the loan still has decades to run.” — TheStreet

Takeaway: Treat 2028 as a planning scenario: if your business depends on AI vendor pricing or tech-sector employment, lock in favorable contract terms now and watch for depreciation schedule changes as an early warning.