Intel Q2 2026 Earnings: A Record Beat, an $11 Billion Paper Loss, and a Rally That's Already Cooling

Intel just did something that almost never happens in the stock market: it beat every major number analysts were watching, by a wide margin, and still reported an $11 billion loss in the same breath. Revenue grew at its fastest pace in fifteen years. Earnings per share came in nearly double what Wall Street expected. And yet, sitting right there in the same press release, was one of the largest quarterly losses in Intel's history. 

The INTC stock fell 2.33% during the regular session on July 23, before results were even out, then jumped as much as 12 to 13% within minutes of the earnings hitting the wire after the market closed, only to give back more than half of that gain by the next morning.

Let's break down what Intel actually reported, why a record quarter can still show up as a loss on paper, and why a beat this large barely moved the needle compared to what Intel's own stock has done before.

Intel Q2 2026 Earnings: Revenue, EPS and Margin Results

Intel released its second-quarter 2026 results after market close on July 23, with a conference call at 5 PM ET, according to Intel's own investor relations release. The numbers cleared every bar Wall Street had set.

Revenue rose 25% year-over-year, Intel's strongest growth rate in more than fifteen years, according to Intel's earnings release and confirmed by CNBC and Reuters. Non-GAAP EPS of $0.42 beat the roughly $0.21-0.22 consensus by more than 90%, swinging from a small non-GAAP loss a year earlier. Non-GAAP gross margin landed at 41.8%, about 280 basis points above Intel's own guidance. 

The one number that looks alarming, an $11 billion GAAP net loss, has a very specific and, once explained, not very scary story behind it. More on that shortly.

Intel Q2 2026 Segment Results: Data Center, Client and Foundry

The growth wasn't evenly spread. One business did most of the heavy lifting.

Data Center and AI, essentially Intel's server chip business, posted its strongest year-over-year growth on record, driven by demand for the newer Xeon 6 processor line, as per Intel's earnings call. That segment alone earned an operating profit of $2.5 billion, a 40% margin.

Intel also quietly renamed its PC division from Client Computing Group to Client Computing and Physical AI Group, a small but telling sign of how central AI-capable hardware has become to Intel's pitch, even for laptops and desktops. Two-thirds of that client revenue now comes from AI PCs. Taken together, AI-linked businesses grew more than 70% year-over-year and made up roughly 70% of Intel's total revenue this quarter.

Intel Foundry, the newer business making chips for outside companies as well as Intel itself, grew revenue but is still deeply unprofitable, an important nuance investors should not gloss over.

The Report Card: Grading Intel Against What Actually Mattered

Heading into this print, the single most useful lens for reading Intel's foundry story wasn't the headline revenue number. It was three specific, narrower numbers, because they are the ones that actually tell you whether Intel's foundry bet is becoming real or staying a story.

External foundry revenue rising from $174 million to $293 million in a single quarter is a genuinely encouraging trend, even though it is still less than 2% of Intel's total $16.1 billion in revenue. The foundry's operating loss narrowing by $348 million quarter-over-quarter, from $2.4 billion to $2.1 billion, is the kind of steady, unglamorous progress that turnarounds are actually made of. What's still missing is the single big, named customer on Intel's newest processes. 

Fortinet, announced July 21 as Intel's first publicly disclosed external customer, is manufacturing its next-generation security chip on the older Intel 4 process at Intel's Fab 34 in Ireland, as per Techtimes and Intel's own announcement. Separately, Intel confirmed an unnamed major cloud provider has committed to 18A production, its newest and most advanced node. Real progress. Just not yet an Apple or Nvidia scale win on the leading edge that would fully settle the debate.

The Strangest Number in the Report: Why Intel "Lost" $11 Billion in Its Best Quarter Ever

Intel's GAAP net loss of $11.0 billion, or $2.16 a share, sounds like disaster. Strip out one specific, unusual accounting item, and Intel actually earned about $2.2 billion for the quarter, comfortably beating expectations. So where did that "loss" come from that has nothing to do with chips, customers, or factories?

It comes from Escrowed Shares, Intel's own term for shares of its stock held in escrow for the US Department of Commerce, tied to the CHIPS Act agreement that gave the government roughly a 9.9% stake in Intel back in 2025, at an entry price of about $20.47 a share. Accounting rules require Intel to revalue this obligation every quarter at the current market price of its own stock. Because Intel's share price has climbed so far above that entry point, the paper value of what Intel technically owes has ballooned, and that increase has to be booked as a loss, a $12.53 billion one this quarter, even though nothing about the underlying business got worse. Quite the opposite.

Think of it like an early employee at a fast-growing startup who was promised a large chunk of ESOPs. If the startup's valuation shoots up before those options are exercised, the company technically owes that employee far more paper value than before, and in certain accounting treatments, that increase shows up as a rising cost on the books, purely because the company is succeeding, not because anything went wrong. Intel's escrowed shares work exactly the same way. The better the stock does, the bigger this specific paper loss gets. It is, in effect, a tax on Intel's own stock price success, and as long as the stock keeps climbing, expect this line to keep showing up, possibly getting larger, every single quarter.

Intel Q3 2026 Guidance: Revenue, EPS, Margins and Capex

Intel didn't just beat Q2. It guided Q3 well above what the Street was modeling.

CFO David Zinsner also raised Intel's 2026 capital expenditure plan from roughly $18 billion to over $20 billion, with 2027 spending expected to rise significantly further, according to TipRanks' live coverage of the call. 

Management framed this as demand-led, not hope-led: Intel disclosed 10 long-term supply agreements with data center customers, some with locked-in pricing, some focused purely on guaranteed volume, and said the company is now supply constrained, meaning customers want more chips than Intel can currently produce, particularly given tight memory and substrate supply. 

It's a bit like trying to book a Tatkal train ticket during festival season: the demand is real and growing, but the number of seats, or in Intel's case, wafers and memory chips, hasn't caught up yet. To help navigate the memory bottleneck specifically, CEO Lip-Bu Tan revealed Intel has hired Seok-hee Lee, the former CEO of SK Hynix.

Why Did Intel Stock Fall, Jump and Then Give Up Its Gains?

To make sense of Thursday, you need to treat it as two separate stock moves, hours apart, driven by completely different things.

The daytime move, down 2.33% to a close of $100.23, had almost nothing to do with Intel's actual business. Alphabet and Tesla had both reported results earlier in the week showing negative free cash flow due to enormous AI-related capital spending, which weighed on AI-linked stocks broadly on a day US indices fell across the board, as per market commentary from Odaily and Bloomberg. Separately, a Semafor report suggested Intel might structure its unfinished Ohio semiconductor campus as an "operating partnership" rather than sell it outright, a day after an earlier, since-denied rumor that SK Hynix might buy the site, as per TipRanks and TheStreet. Neither story touched Intel's actual quarter, but both added to a month where the stock had already fallen roughly a quarter from its June 30 all-time high of $142.35, largely on profit-taking after an extraordinary run.

The after-hours move was a reaction to real numbers. Within minutes of the 4 PM ET release, Intel shares jumped as much as 12 to 13%, briefly touching levels above $112, as traders digested the revenue beat, the EPS beat, and Q3 guidance that blew past consensus, according to Reuters and Google Finance data. By roughly 7 PM ET the same evening, that gain had already cooled to under 4%. That kind of fade is not unusual for Intel. It is, in fact, the single most useful thing to understand about how this particular stock behaves around earnings.

The Half-Life of a Beat: How Intel Stock Has Reacted to Recent Earnings Reports

Here's a pattern worth understanding if you plan to hold Intel through future earnings prints: the size of a beat and the size of the stock's reaction have stopped moving together.

In April 2026, a strong Q1 beat sent Intel's stock up 23.6% in a single session, one of its best days in decades. This time, an even larger revenue beat and a far larger EPS beat produced a smaller, and quickly fading, reaction. Why would a bigger beat get a smaller response? Because Intel's stock had already climbed roughly 170% for the year, and well over 300% across the past twelve months, heading into this report, largely on the expectation that a turnaround like this was coming. A stock that has already priced in good news doesn't get rewarded twice for confirming it.

It's a bit like a topper who scores 95% in every single exam. The first time, the whole class is stunned. By the tenth time, another 95% barely gets a reaction, even though nothing about the performance changed. Good news that arrives right on schedule eventually stops being news. The lesson isn't that this quarter didn't matter. It's that with a stock this richly re-rated, the size of a headline beat is no longer a reliable guide to how it trades in the days that follow.

What Intel Q2 2026 Earnings Mean for Investors

The bull case strengthens if external foundry revenue keeps climbing toward the $400-500 million range over the next couple of quarters, the Foundry operating loss keeps narrowing at a similar pace, at least one more marquee customer commits to 18A or 14A capacity in real volume, and the newly raised capex converts into revenue rather than cost overruns. 

The bear case strengthens if gross margin compresses from memory and substrate cost inflation, which Zinsner has already flagged as a live risk into the second half, if the "subseasonal" PC market he described actually weighs on Client Computing, if the Ohio campus sees further execution slips regardless of its final ownership structure, or if today's supply constraints turn out to reflect years of under-investment Intel still hasn't fully closed.

One thing worth remembering regardless of which case you lean toward: as long as Intel's stock keeps climbing, that escrowed-shares GAAP charge is likely to recur, or even grow, every quarter, for purely mechanical reasons unrelated to day-to-day performance. If you're using Intel's headline GAAP profit or loss as your main signal quarter to quarter, you're reading the wrong line. The non-GAAP numbers, and the segment trends underneath them, are where the real story lives.

Our Take

Strip away the noise, and this was a genuinely strong quarter. Revenue grew at its fastest pace in fifteen years. The Data Center and AI business is growing at a rate that would have seemed implausible twelve months ago. The Foundry business, still loss-making, is losing less each quarter while its external revenue climbs. Management raised capex because real customer demand is asking for it, not because of optimism alone.

What this quarter wasn't, was a print that resolves the debate over whether Intel's turnaround is already priced into a stock that gained more than 300% over the past year. A beat this size would have been front-page news at almost any earlier point in Intel's recent history. This time, it barely held the market's attention for a full evening. That gap, between the genuine quality of the results and how briefly the market rewarded them, is the most important thing this earnings report actually tells you, more than any single line item in the press release.