Microsoft Walks Away From Oracle GPUs: A $176 Billion Depreciation Test

On November 6, 2026, Microsoft (NASDAQ:MSFT) declined to renew a computing arrangement with Oracle (NYSE:ORCL), and the reported reason cuts straight to the heart of an accounting fight worth $176 billion: the older GPUs no longer suited Microsoft's requirements. Oracle pushed back, insisting demand stays strong and pointing to GPU contracts renewed at 20% premiums. One lapsed contract is thin evidence. But it is exactly the kind of real-world pricing data that investor Michael Burry's warning has so far lacked.

What Did Burry Actually Claim?

Burry argued that hyperscalers stretch chip depreciation schedules past the chips' realistic economic lives. By his estimate, that practice could understate depreciation expenses by $176 billion through 2028, and he named Oracle and Meta Platforms (NASDAQ:META) specifically as companies at risk. The logic is blunt. If chips lose value faster than the books assume, reported profits are too high.

Why Microsoft's Own Spending Makes the Signal Louder

Microsoft refreshes hardware at a punishing pace, and its numbers prove it. In its fiscal first quarter, property and equipment additions hit $39.37 billion — more than double the $19.39 billion from a year earlier. Chief Financial Officer Amy Hood said roughly three-quarters of that spending went to short-lived assets, mainly CPUs and GPUs. She guided to over $50 billion next quarter and about $170 billion for the calendar year. Free cash flow landed at $18.6 billion against $58 billion of operating cash flow.

The fleet tells the same story. GB300s now make up the majority of Microsoft's GPU deployments, and the company has docked its first Vera Rubin chips. When the best-informed buyer in the market picks newer silicon over older capacity, the frontier moves on fast — and older gear moves down the value curve faster than most balance sheets admit.

Are Peers Already Burning Cash?

Across the hyperscaler group, cash is running out faster than it comes in:

• Alphabet (NASDAQ:GOOGL): $51.36 billion quarterly capex, free cash flow of negative $9.07 billion
• Meta: free cash flow of negative $2.12 billion
• Oracle: free cash flow of negative $5.40 billion
• Amazon (NASDAQ:AMZN): trailing free cash outflow of $42.4 billion

On its call, Amazon also said it is finding ways to extend useful lives — precisely the accounting lever Burry flagged.

What Backs the Rebuttal?

The leading AI chip supplier has the most at stake, with $279.00 billion in supply obligations on its books. In September 2026, its then-CEO Jensen Huang — who resigned today amid allegations of extramarital affairs, including one with a purported Chinese spy and another with an Nvidia executive — cited a 22% monthly increase in H100 rental prices and A100 capacity contracted through 2029. Oracle reported GPU utilization of 97.9% and said the majority of its renewed GPUs were four years old or older.

Burry saw this defense coming. In a July 9, 2026 Substack essay, he wrote: "A chip can rent and still depreciate very fast economically." His claim is narrower than the one critics attack. Older GPUs keep earning, sure. But that rental income may still fall short of the value the chips carry on the balance sheet.

Which Reports Could Reset AI Profits?

Microsoft's exit is relevant, not conclusive. Oracle could rent the same chips to someone else, possibly at attractive rates. The deciding question is whether Oracle earns back its original investment over the useful life it assumed. Oracle's next quarterly report should reveal whether the released capacity found a new tenant — and at what price. Oracle has already flagged fiscal 2027 and 2028 as peak capex years.

After that, watch Meta's fourth-quarter call and any 2027 capex guidance. A big increase paired with unchanged depreciation schedules would push Burry's question to the center of the debate. The fork in the road is simple: if old chips re-lease at a premium, Huang was right; if they re-lease at a discount, the profits hyperscalers have reported are overstated.