Intel Lost $11 Billion in Its Best Quarter in 15 Years. The Charge Is Its Own Stock Price

On July 23, 2026, Intel reported second-quarter revenue of $16.1 billion, up 25% year over year. It was the company's fastest quarterly growth since 2011, and it beat Wall Street's $14.4 billion estimate by about 12%. The same release reported a GAAP net loss of $11.0 billion, or $2.16 a share. Both statements are accurate, and one line reconciles them.

The story works until this line, buried among the non-operating items: a mark-to-market loss on Escrowed Shares of $12.5 billion, bigger than any product category Intel sells.

Here is what the escrowed shares are, because the name hides the plot. In August 2025 the company and the Trump administration announced that the U.S. government would invest $8.9 billion in Intel common stock — 433.3 million shares at $20.47 each, roughly 9.9% of the company. The money was not new: $5.7 billion came from the remaining unpaid CHIPS Act grants and $3.2 billion from the Secure Enclave program, converting money that was scheduled to be a subsidy into an ownership stake. Intel received about $5.7 billion of it in cash early, with prior project milestones removed. The agreement converted scheduled grant money into equity and eliminated the earlier grants' claw-back and profit-sharing provisions, replaced by a passive government holding that agrees to vote with Intel's board.

The accounting is the unusual part. Because the government's shares are held in escrow — restricted, released as conditions are met — Intel does not record the stake as an ordinary issuance of stock. It carries the escrowed shares as a liability measured at fair value, and that fair value is, in effect, the market price of Intel's own common stock. At closing Intel allocated $5.7 billion of cash proceeds against the escrowed shares at a fair value of $3.9 billion. Every quarter thereafter, the liability is re-measured. When Intel's share price rises, the liability rises, and the increase is charged to income as a loss. When Intel's stock falls, the item reverses into a gain.

Now follow the same dollar into the income statement's timing. INTC spent the spring of 2026 climbing toward a June peak near $142, and the quarterly charge tracked the climb. Divide the $12.5 billion by the 433 million escrowed shares and you get roughly $29 per share of fair-value movement — the same order of magnitude as the rally itself moved. In the prior quarter the same line cost about $1 billion when Intel reported a $3.7 billion GAAP net loss. In Q2, with the stock's biggest legs of the move, the escrow wrote a $12.5 billion check. Intel's own disclosure says the charge represents the net change in fair value of escrowed shares both released during the quarter and still held at quarter end.

The benign explanation deserves its full hearing, because on inspection it clears most of the worry. The loss is non-cash. It sits below the operating line, in non-operating items, so operating income can turn positive while net income collapses. Non-GAAP EPS of $0.42 — nearly double the $0.22 analysts expected — excludes it entirely, and the market reacted to the adjusted number, sending the stock up about 11% in extended trading. If the stock were to fall next quarter, the line would print a gain. Nothing here is misstated, hidden, or missing from the disclosure. Management did not mislead investors; the reconciliation is in the release.

But the clearing record changes where the cost lands rather than erasing it. This is the part of the ledger the adjusted number does not show.

First, the dilution. Those shares actually go to the government as they release. At today's price near $90, the stake is worth roughly $39 billion; near the June peak it was above $60 billion — against an $8.9 billion commitment. The difference does not evaporate because it is booked as an accounting loss; it settles as new shares landing in government hands. That is permanent dilution of everyone who already owns Intel, and under the deal terms the government also holds a warrant on an additional 5% of common shares if the company ever owns less than half of its own foundry business. The quarterly "loss" is the meter running on a transfer of ownership.

Second, the cash. Intel reported $7.0 billion of operating cash flow for Q2, up from $2.1 billion a year earlier, and $8.1 billion for the first half. Look behind the number and most of it is bookkeeping returning to itself: first-half results add back $13.6 billion of escrowed-share marks and $4.0 billion of restructuring charges against a GAAP net loss of roughly $15 billion. The government's early cash arrived the same way — money in hand today that the company will repay in equity. By Intel's own definition of adjusted free cash flow, which nets capital spending, partner contributions, and finance leases, the quarter was a negative $8.4 billion, worse than the negative $1.1 billion a year earlier. This is not a company manufacturing profit from nothing; it is a company explaining why cash on hand is not the same as earnings, and vice versa.

The revenue, to be fair, is real and accelerating. Data Center and AI revenue rose 59% to $6.3 billion, client computing rose 13%, and the foundry segment grew 31% to $5.8 billion — though it still lost $2.1 billion in the quarter, and its largest customer remains Intel itself. Intel guided third-quarter revenue of $15.8 billion to $16.8 billion, above expectations. The operating parts of the story are genuinely improving.

That is what makes the stock's position uncomfortable, and it gets to the phrase attached to this situation: paying now for revenue due later. At about $89.60 in late August, Intel carries a market value near $474 billion — roughly 8.3 times trailing revenue and about 40 times trailing EBITDA, on a company that still loses money under GAAP. Since the start of the year the stock is up roughly 143%, and over the past twelve months about 275%, after more than tripling off last year's lows. The revenue that must eventually justify that price is mostly due later. The foundry's confirmed external marquee customer on 18A remains Microsoft; reports of Apple, AMD, Google, and Nvidia weighing deals are conversations, not contracts. The momentum in AI server CPUs is real but young. A $474 billion valuation is the market writing the check today for revenue the company has not yet earned.

Here is the shareholder invoice, whatever the stock does next. If the stock keeps rising, every quarter books another large GAAP loss, the value accumulating inside the government's stake grows, and the eventual share releases dilute existing holders further — the reported loss is cosmetic, the dilution is not. If the stock falls, the line reverses into gains and the income statements look cleaner, but only because the enthusiasm that supports the valuation cooled. In neither case is there a material-misstatement problem: the mechanism is disclosed in the deal documents, the release, and the 10-Q, and it is corroborated by arithmetic anyone can reproduce. The genuine exposure is not accounting — it is paying a $474 billion price for foundry and AI revenue that has to arrive later, while the very rally that justifies the price keeps converting into reported losses and dilution on the way there.

The settling events to watch, in order: the next quarterly report, where the escrow line keeps mirroring the stock's direction but the number that matters is whether foundry external revenue and data-center momentum grow faster than the launch; and any disclosure of escrow releases, which will quantify the dilution already permanent. The escrow line tells you what the market believes about this stock. The foundry's external customer list tells you what Intel has actually sold.