By Intel reported an $11.0bn net loss in its second quarter of 2026 and the business was not losing money. Those two statements are both true, and the reconciliation between them is the single most misunderstood number in large-cap semiconductors. Intel closed at $95.80 on 4 September 2026, up 4.51%, having risen roughly 289% over twelve months from a low of $24.08. The loss that dominates every summary of the company is driven by a $13.6bn non-cash mark-to-market charge on Escrowed Shares – Intel stock held in escrow for the US Department of Commerce – and that charge is measured in Intel’s own equity. It therefore gets larger when the share price goes up. Intel is the only company of its size where a rally mechanically manufactures a bigger reported loss.
This is not a technicality worth a footnote. It inverts the normal relationship between share price and reported earnings, and it will keep doing so. Intel traded at $87.48 in late August. It is $95.80 now. Everything else equal, that 9.5% appreciation makes the next quarter’s escrow charge bigger than the last one, which means the next set of headlines will describe a larger GAAP loss produced in part by the stock having gone up. Anyone screening on reported earnings, and any model that treats the loss as an operating result, is reading Intel’s equity appreciation as though it were business deterioration. The figure sits in plain sight in the company’s own cash-flow reconciliation, added straight back as a non-cash item.
Key facts
Intel closed at $95.80 on 4 September 2026, up 4.51%, and 32.0% below its 52-week closing high of $140.94 – stockanalysis.com daily closes, retrieved 5 September 2026
Second-quarter revenue was $16.13bn, up from $13.65bn a year earlier, with gross profit of $6.51bn for a gross margin of 40.4% – Intel Q2 2026 results, 23 July 2026 (margin is a FinanceFeeds calculation)
GAAP net loss was $11.03bn, of which $13.62bn is a non-cash mark-to-market loss on Escrowed Shares, listed as an add-back in the cash-flow reconciliation against nil in the prior-year period – same filing
Restructuring and other charges were $3.97bn in the period, against $382m a year earlier – same filing
Research and development ran $3.37bn, equal to 20.9% of revenue – Intel Q2 2026 Form 10-Q, filed 24 July 2026 (FinanceFeeds calculation)
Intel has approximately 5,044 million shares outstanding, for a market capitalisation of roughly $483bn at the 4 September close – Q2 2026 Form 10-Q cover, FinanceFeeds calculation
The company has removed roughly 39,700 people in two years – see our analysis of Intel’s headcount reduction
What the escrow charge actually is
Under Intel’s arrangements with the US government tied to CHIPS Act support, a block of Intel shares sits in escrow for release to the Department of Commerce. Because the obligation is settled in Intel’s own stock and carried at fair value, it behaves as a liability that tracks the share price. When Intel’s equity appreciates, the value of what Intel owes appreciates with it, and the increase runs through the income statement as a loss.
The mechanics are visible in the primary filing rather than inferred. In the reconciliation from net loss to cash provided by operating activities, Intel adds back a line reading “Mark-to-market (gains) losses on Escrowed Shares” of 13,619, in millions of dollars, against a blank for the same period a year earlier. An add-back in that statement is the accounting definition of a charge that consumed no cash.
Two consequences follow. The first is that Intel’s GAAP earnings are, for now, an unreliable guide to the operating business, and the gap is not small or seasonal – it is larger than the loss itself. The second is directional and easy to miss: because the charge scales with the share price, the better the equity performs, the worse the reported number looks. That relationship holds until the escrowed shares are released, and it means Intel’s headline results and its shareholder outcomes will keep pointing in opposite directions.
None of this makes the operating business good. It makes it separately assessable, which is the necessary first step.
What the operating business actually did
Revenue of $16.13bn was up from $13.65bn a year earlier, and management framed the quarter as an execution beat. “We delivered a strong second quarter, exceeding our financial guidance on robust demand and improved execution, including volume upside driven by higher factory yields and improved cycle times,” said Dave Zinsner, Intel chief financial officer, in the results release.
Higher factory yields is the phrase that matters for a company whose central problem for half a decade has been manufacturing. Yield is the variable that converts Intel’s enormous fixed-cost base from a liability into leverage, because the fabs are already built and paid for. Gross margin of 40.4% remains far below the company’s historical peak and far below what a healthy foundry-plus-products business should generate, but the direction is the argument.
Lip-Bu Tan, Intel chief executive, positioned the company against the demand cycle rather than against a specific competitor: “AI is driving unprecedented demand for compute, and as we continue to execute, Intel is well-positioned to capture sustainable growth across our CPU franchise, ASICs, advanced packaging and vast wafer foundry network.”
Note what is being sold there. Not a GPU roadmap. A CPU franchise, custom ASIC work, advanced packaging and foundry capacity. That is a deliberate repositioning away from competing head-on with Nvidia on accelerators and toward being infrastructure that other designers use. Whether the market pays for that is the entire Intel question.
Zinsner also flagged the cost of it: Intel is “meaningfully increasing our investments in equipment, clean room space, and substrates” to support expected growth this year and next. Restructuring and other charges of $3.97bn in the quarter, against $382m a year earlier, show the other side of the same reshaping.
Three companies, one cycle, three positions in the value chain
Intel’s 40.4% gross margin is best understood next to its peers this quarter, because the three of them map the AI supply chain from end to end.
CompanyLatest gross marginPosition in the chain
Micron84.6%Supply-constrained memory; sets price
AMD54% GAAPFabless designer; pays memory, pays TSMC
Intel40.4%Designs and manufactures; carries the fabs
The pattern is not about competence. It is about scarcity. Micron earns the highest margin because memory supply is the binding constraint in this cycle and only three companies can relieve it. AMD sits in the middle, capturing design value while paying the memory tax. Intel earns the least gross margin of the three and is the only one that must also fund fabrication out of that margin.
That is the bear case in one table, and it is also the bull case. Owning the fabs is what compresses Intel’s margin today and what makes it the only Western company that could supply leading-edge capacity to everyone else if the industry decides concentration in Taiwan is a risk it can no longer carry. The same asset is the drag and the option.
The depreciation line is the real constraint
There is a number in Intel’s cash-flow statement that explains the gross margin better than any commentary about competitiveness. Depreciation in the quarter was $5.89bn. Against revenue of $16.13bn, that is 36.5% of every dollar of revenue consumed by the amortised cost of plant Intel has already built (FinanceFeeds calculation from the Q2 2026 results).
That single line is the difference between Intel and a fabless designer. AMD’s cost base scales with what it buys from TSMC; if demand falls, its purchases fall. Intel’s does not. The fabs depreciate on schedule whether or not wafers move through them, which is why utilisation, not pricing, is the dominant variable in Intel’s margin and why yield improvements translate so directly into profit.
It also explains why the company can post a 40.4% gross margin and still generate substantial operating cash flow: depreciation is a non-cash charge, so a large share of what depresses reported margin is added back before cash is measured. Intel’s cash generation has always looked materially better than its earnings, and in a period when its earnings also carry a $13.6bn non-cash equity charge, the divergence between the two is about as wide as it has ever been for a company this size.
The forward risk is that Zinsner’s “meaningfully increasing our investments in equipment, clean room space, and substrates” adds to that fixed base. New capacity raises future depreciation before it raises revenue. If the AI demand Tan describes arrives on schedule, that timing gap is a rounding error. If it slips by a year, the depreciation lands first and the margin gets worse before it gets better.
The sovereign shareholder changes the risk shape
Intel is now partly owned by the state that regulates it, and that is a genuinely different kind of security to analyse. A government shareholder is a backstop against the disaster scenario, which is why Intel’s downside is arguably better protected than any peer’s. It is also a constituency with objectives that are not shareholder returns: domestic capacity, employment, supply-chain security.
Those objectives usually argue for building more fabs in more places than a purely commercial operator would choose, and for building them on a timetable set by politics. Investors buying Intel for the foundry option are buying a business whose capital allocation has a second decision-maker. In a downturn that is protection. In an upturn it is a ceiling on returns on capital.
The escrow charge is the visible accounting expression of that relationship, and it will remain in the numbers until the shares are released.
What happens next
First, expect the GAAP loss to persist or grow while the stock does well. This is arithmetic, not forecasting. If Intel’s shares appreciate over the next quarter, the escrow charge grows and the reported loss widens. The signal to watch is non-GAAP earnings and operating cash flow, and the risk is that headline-driven selling on a “wider loss” print creates a gap between the reported number and the business.
Second, gross margin is the scoreboard for the manufacturing turnaround. Zinsner attributed the quarter’s upside to higher factory yields and improved cycle times. If that is durable, it shows up as gross margin climbing from 40.4% toward the mid-forties over the coming quarters, with no change in revenue required. If margin stalls near 40% while capital spending rises, the turnaround thesis weakens regardless of what revenue does.
Third, the foundry proof point is an external customer at leading edge. Tan’s framing rests on a “vast wafer foundry network” serving other designers. The market will not pay for that until a named, high-volume external customer commits publicly. Until then Intel is valued as a products company with a very expensive manufacturing arm, and the fabs are counted as a cost rather than an asset.
Our numbers: bull $150, base $112, bear $55, against a spot of $95.80. What would change our mind on the bull case is a second consecutive quarter of flat gross margin alongside rising capital expenditure, which would suggest the yield improvement was a mix effect rather than a process win. What would change our mind on the bear case is a leading-edge foundry commitment from a customer with real volume, which would re-rate the fabs from cost centre to strategic asset overnight. For the manufacturing benchmark Intel is measured against, our TSMC analysis sets out what a fully utilised leading-edge foundry actually earns.
Frequently asked questions
Why did Intel report an $11bn loss if the business is improving?
Because $13.62bn of non-cash mark-to-market loss on Escrowed Shares runs through the income statement. The shares are held in escrow for the US Department of Commerce and are carried at fair value, so the obligation grows with Intel’s share price. It appears as an add-back in Intel’s cash-flow reconciliation, which is the accounting confirmation that it consumed no cash.
Does the escrow charge get bigger if Intel stock rises?
Yes. Because the liability is measured in Intel’s own equity, appreciation in the share price increases the value of what Intel owes and produces a larger charge. This is why Intel’s reported results and its shareholder outcomes can move in opposite directions, and it persists until the escrowed shares are released.
What is Intel’s gross margin and how does it compare?
Gross margin was 40.4% in the second quarter of 2026, calculated from gross profit of $6.51bn on revenue of $16.13bn. That compares with 54% GAAP at AMD and 84.6% at Micron in their most recent quarters. Intel earns the least of the three and is the only one funding leading-edge fabrication from that margin.
What would make Intel’s foundry business worth something?
A named external customer committing high volume at the leading edge. Chief executive Lip-Bu Tan describes a “vast wafer foundry network” as central to the strategy, but the market currently prices Intel’s fabs as a cost rather than an asset, and only a credible third-party commitment changes that.
How much has Intel stock risen over the past year?
Intel closed at $95.80 on 4 September 2026 against a 52-week low of $24.08, an increase of roughly 289%. The shares nonetheless remain 32.0% below their 52-week closing high of $140.94.
This article is analysis and information, not investment advice. Scenario levels are the author’s estimates based on company filings and are not price targets or recommendations. Trading and investing carry risk, including the total loss of capital. Figures were verified against primary sources on 5 September 2026 and may have moved since.