The figures in Intel's earnings appear contradictory at first glance. Revenue grew 25% year-over-year, marking the strongest growth rate in more than 15 years, and operating income swung from a loss a year earlier into a profit. Despite this, the bottom line showed a net loss of $11.033 billion, more than triple the loss recorded in the same quarter a year ago. This gap did not stem from a business slowdown but from the accounting treatment of escrow shares tied to the US Department of Commerce's equity stake acquired in August 2025. The higher Intel's stock price climbs, the larger the paper loss on Intel's books grows—and on the other side of that same coin, the unrealized gain on the shares the US government holds swells in tandem.
Revenue Up 25%, Yet a $11 Billion Net Loss: The Paradox in Intel's Earnings
Intel's second-quarter 2026 earnings, released on July 23, 2026, showed revenue of $16.1 billion, up 25% from the same period a year earlier. That growth rate is the company's strongest in more than 15 years. GAAP operating income came in at $1.796 billion, a turnaround from an operating loss of $3.176 billion in the same quarter last year. Gross margin under GAAP also improved by 12.9 percentage points year-over-year to 40.4%.
Yet the bottom line showed a GAAP net loss of $11.033 billion, with a loss per share (EPS) of $2.16. Compared with the net loss of $2.918 billion in the same period last year, the deficit more than tripled. The single largest factor behind this gap was a $12.529 billion mark-to-market loss on escrow shares tied to the equity stake held by the US Department of Commerce. Even as the company turned profitable at the operating-income level, this one line item alone drove the company to an $11 billion net loss.
Meanwhile, non-GAAP EPS, which excludes special items, came in at $0.42—nearly double analysts' consensus estimate of $0.21 to $0.22. On a GAAP basis, Intel has now posted losses for three consecutive quarters: a $591 million loss in the October–December 2025 quarter, a $3.728 billion loss in the January–March 2026 quarter, and now the $11.033 billion loss in the April–June quarter. However, the nature of each loss differs by quarter. The January–March loss was driven mainly by $4.070 billion in restructuring-related charges, including a goodwill impairment tied to the Mobileye business, whereas the April–June loss centered on the non-cash mark-to-market charge on escrow shares. As discussed below, the three core segments—the data center and AI business, the foundry business, and the client business—each posted either revenue growth or profitability, suggesting that the turnaround in Intel's core operations is broadly on track.
Why the Escrow Share Valuation Loss Arises: The Mechanics Behind the August 2025 US Government Investment
Intel's earnings materials attribute the primary driver of the net loss to "mark-to-market losses on escrow shares." Escrow shares refer to Intel's own shares that the company temporarily holds under an agreement with the US Department of Commerce, transferring them to the department incrementally as certain conditions are met. Under accounting rules, Intel must remeasure this transfer obligation as a liability at fair value—meaning that as Intel's stock price rises, the value of the shares to be transferred grows, increasing the liability, with that increase recognized directly as an expense.
Here is the background. In August 2025, the US Department of Commerce agreed to take an equity stake in Intel, funded by $5.7 billion in unpaid grants under the US CHIPS Act—which provides subsidies for the semiconductor industry—and $3.2 billion from the Secure Enclave program, aimed at securing domestic production of defense-related semiconductors. Of this, the CHIPS Act portion, amounting to 274,583,000 shares (at $20.74 per share), has already been directly issued.
The Secure Enclave portion, meanwhile, consists of 158,740,000 shares (at $20.00 per share) held in an escrow account, to be transferred to the department incrementally each time Intel receives cash. As of June 27, 2026, approximately 143 million shares remain untransferred. This entire pool of escrow shares, including the untransferred portion, is subject to quarterly mark-to-market valuation. Separately, warrants covering up to 240,516,150 shares (exercise price of $20) have also been issued, exercisable only if Intel divests a majority stake in its foundry business—but as of now, the exercise conditions have not been met.
Before the Q2 earnings announcement in July 2026, Intel shares had risen 171.6% year-to-date, climbing from a 2025 year-end closing price of $36.90 to $100.23. This stock price surge inflated the valuation of the escrow shares, resulting in a $12.529 billion mark-to-market loss recognized in the April–June quarter alone.
The same stock price increase has also boosted the value of the 274,583,000 shares already directly issued to the US government. Based on an acquisition cost of $20.74 per share (roughly $5.7 billion) and valued at the July 23, 2026 closing price of $100.23, the holding's market value comes to approximately $27.5 billion, implying an unrealized gain of roughly $21.8 billion. While the mark-to-market loss on escrow shares and the unrealized gain on the government's already-held stake are recorded as separate accounting items, both stem from the same underlying rise in Intel's share price.
It's also worth noting that this valuation loss involves no cash movement. Even as the market value of the escrow shares rises, no cash actually flows from Intel to the Department of Commerce. The company is simply recognizing, per accounting rules, the increase in the theoretical value of shares it will transfer in the future. Conversely, if Intel's stock price falls, the value of the escrow shares would decline as well, opening the door for the company to record a valuation gain in a future quarter.
Strong DCAI, a Narrowing Foundry Loss, and the Client Business Renamed CCPG
| Segment | Revenue | YoY Change | Operating Income/Loss |
|---|---|---|---|
| DCAI (Data Center and AI) | $6.26 billion | +59% | $2.47 billion profit |
| Intel Foundry | $5.77 billion | +31% | $2.09 billion loss (Q1 loss was $2.4 billion) |
| CCPG (formerly CCG) | $8.88 billion | — | $2.34 billion profit |
DCAI (Data Center and AI) revenue rose 59% year-over-year to $6.26 billion, a sharp acceleration from the 22% year-over-year growth seen in the January–March quarter. AI-driven server demand was the main growth driver. CEO Tan stated in the earnings materials, "AI is driving unprecedented compute demand, and Intel is well positioned to deliver sustained growth through our CPU product lineup, ASICs, advanced packaging, and wafer foundry network."
The Intel Foundry segment as a whole posted revenue of $5.77 billion against an operating loss of $2.09 billion. While the bulk of the revenue still comes from internal customers, the loss narrowed by roughly 13% from the $2.4 billion loss posted in the January–March quarter.
CCPG (formerly CCG, now the Client Computing and Physical AI Group), centered on client CPU business, posted revenue of $8.88 billion and operating income of $2.34 billion. The renaming from CCG (Client Computing Group) to CCPG signals the company's intent to fold robotics and physical AI into the business. Among the three segments, this remains the most stable source of earnings, serving as the backbone that offsets the volatility in DCAI and Foundry. This renaming is also part of the broader business restructuring that CEO Lip-Bu Tan has pursued since taking office. Moving away from a business structure reliant solely on client CPUs, the name change signals in advance the company's intent to capture new demand from robotics and physical AI.
The Fate of Two External 14A Customers, and Improved Profitability with a Workforce of 82,300
On the manufacturing side, Intel Foundry has moved its next-generation process, Intel 18A-P, into risk production (trial production just ahead of full mass production). Separately, the company has also launched a new data center CPU, Xeon 6+ (code-named Clearwater Forest), manufactured on the standard Intel 18A process—positioned as the first server-oriented product of the Intel 18A generation. CFO Zinsner noted that "improved factory yields and shorter cycle times drove the upside in volumes," and outlined plans to sharply increase investment in equipment, cleanrooms, and substrates to meet AI-driven demand growth.
Looking further ahead to Intel 14A, the company had disclosed as of January 2026 that two prospective external customers were conducting early PDK (Process Design Kit, the standard set of data needed for chip design) evaluations. The Q2 earnings materials likewise indicate that discussions with external customers are ongoing as the PDK is refined further. Whether these customers make initial design commitments is expected to become clear sometime between late 2026 and early 2027. Whether that leads to formal mass-production investment decisions will significantly shape the scale of Intel Foundry's external revenue.
On the workforce front, headcount stood at 82,300, down about 19% from 101,400 a year earlier. This decline includes roughly 3,000 employees who were deconsolidated following the sale of a 51% stake in Altera, but the bulk of the reduction stems from headcount cuts that Tan has pursued since becoming CEO in March 2025. Dividing revenue of $16.1 billion by the headcount of 82,300 yields revenue per employee of approximately $196,000. In the same period a year earlier, revenue was roughly $12.88 billion (calculated by working backward from the 25% year-over-year growth rate), which, divided by the then-headcount of 101,400, comes to about $127,000 per employee—implying a 54% increase.
The workforce reduction is not limited to this particular quarter. Headcount, which stood at roughly 132,000 in 2022, has fallen into the low 80,000s through multiple rounds of restructuring since then. The Tan administration is carrying forward and accelerating this trend.
Q3 Guidance, Stock Price Swings, and What the Earnings Materials Left Unsaid
Intel's company guidance for the July–September 2026 quarter (Q3) calls for revenue of $15.8 billion to $16.8 billion, GAAP EPS of $0.31, and non-GAAP EPS of $0.38. If achieved, this would extend the recovery momentum in the core business seen in the April–June quarter. The stock market's reaction was volatile. In after-hours trading immediately following the earnings release, Intel shares briefly jumped 12–13%, but by the next morning's session, more than half of that gain had evaporated. Ahead of the earnings release, Intel shares had already risen 171.6% year-to-date, suggesting that much of the good news may already have been priced in.
For Japanese readers, the point of relevance lies in the yen-converted figures. The net loss of $11.033 billion translates to roughly ¥1.8 trillion at the July 23, 2026 exchange rate of ¥163 to the dollar. If Intel Foundry expands its capital expenditures for 18A and 14A, domestic semiconductor equipment makers such as Tokyo Electron could stand to benefit.
The earnings materials do not disclose the names of the two prospective external customers for Intel 14A, nor a specific timeline for mass-production investment decisions. Nor do they address the flip side of the risk on the escrow share valuation loss—namely, the potential for it to reverse into a valuation gain if the stock price declines. These are two points investors will need to track independently in the coming quarters.
If Intel succeeds in landing external 14A customers, and if the mark-to-market loss on escrow shares begins to shrink as the stock price reverses, the company would emerge from the current run of quarters in which GAAP and non-GAAP figures appear twisted against each other. Until then, the underlying strength of the core business—reflected in revenue growth and a return to operating profit—will continue to coexist with the appearance of an accounting loss stemming from the government's equity stake.