Intel Beats Q2 Estimates and Raises Capital-Spending Plans; Shares Fall After Initial Gain

Beat and Fell in the Same Breath

Intel closed the books on its second quarter on Thursday, July 23, 2026, and the numbers were hard to argue with. Revenue came in at $16.13 billion, up 25% from a year earlier and $1.8 billion above what the company itself had predicted^[1][6]. Adjusted earnings hit $0.42 a share, more than double the $0.20 Intel had guided investors to expect^[1][6]. It was the seventh straight quarter Intel beat its own forecast^[5].

Data center and AI chips drove it, with that business growing 59% year over year^[5][6]. Wall Street had expected less than what Intel delivered — analysts were modeling around $14.33 billion in revenue^[8].

Then Intel said what it planned to spend to keep that going. The company raised its 2026 capital budget to more than $20 billion, up from about $18 billion, and its chief financial officer, David Zinsner, told investors 2027 spending would be "significantly above" that^[1][7]. The stock jumped after hours on the earnings beat. Then it gave the gain back and fell in the next trading session^[8][9]. Outlets don't even agree on how big that after-hours jump was — one report put it above 13%, others lower — which is itself a sign of how fast the story moved^[9].

Both halves of that story are true at once: a record quarter, and a stock that dropped anyway. That's the whole dispute in miniature. Nobody disagrees Intel is spending more. What they disagree on is whether that's the smart move or the risky one.

The $11 Billion Loss That Wasn't

Buried in the same report is a number that looks far worse than anything above: a GAAP net loss of $11.03 billion, or $2.16 a share^[3][4]. Read on its own, that looks like a company in serious trouble. It isn't, and the reason why says a lot about who else has a stake in Intel's stock price.

In August 2025, Intel struck a deal with the Trump administration that gave the U.S. government an equity stake in the company, roughly 10%, tied to earlier CHIPS Act funding^[15][17][18]. Part of that arrangement involves shares held in escrow that Intel owes the Commerce Department. Accounting rules require Intel to revalue that obligation every quarter, based on Intel's own stock price. When Intel's stock rose sharply this year, the shares it owes the government became more valuable — and Intel had to book that increase as a $12,529 million non-cash loss^[3][4]. No cash left the building. It's an IOU getting more expensive on paper because the underlying stock did well.

Add a separate $3.9 billion non-cash write-down at Mobileye, Intel's self-driving-technology subsidiary, and you get most of the $11 billion^[3][4]. Strip out both one-time items and Intel actually earned $2.2 billion in the quarter, on a non-GAAP basis^[4]. The headline loss and the underlying business are telling two different stories, and only one of them involves real money moving.

Why a Factory Has to Be Built Before It Sells Anything

The more consequential fight is over the capital-spending increase itself, and it turns on a mechanic of the chipmaking business that's easy to miss: in contract manufacturing, you can't sell space you haven't built yet^[22]. A chip factory, called a fab, costs billions of dollars and takes years to construct before it produces a single sellable chip. That means a foundry — a company that manufactures chips designed by other companies — has to build capacity on a bet that customers will show up later, not after they've already signed.

That's the case Intel's management makes for spending more. CEO Lip-Bu Tan told investors demand "continues to outpace" what Intel can currently supply, and said AI-related business grew more than 70% year over year and now makes up about 70% of Intel's revenue^[5]. If a competitor has open capacity and Intel doesn't, a customer who needs chips built now goes elsewhere — and stays there for the life of that product line^[22]. Underbuilding, in this framing, isn't caution. It's forfeiting the contract.

Skeptical investors read the same fact and draw the opposite conclusion, because of what happens to spending after the ribbon-cutting. Money spent on a factory doesn't just disappear from the books once it's spent — it gets spread out over many future years as depreciation, an accounting charge that lands inside the cost of making each chip. A fab built in 2027 will be dragging down Intel's profit margins in 2029 and 2030, whether that fab is running at full capacity or sitting half-empty^[9][21]. Meanwhile the cash goes out today, for orders that may or may not materialize. That combination — cash out now, a margin hit for years, on a bet not yet proven — is exactly what a Bank of America survey of fund managers has been picking up: a net 20% said in November 2025 that companies were investing too aggressively, the first such reading in about two decades, and by February 2026 that had climbed to a record 35%, the highest share in the survey's history^[19][23]. Companies that raised AI spending plans this year have generally been punished for it in their stock price^[8].

A third camp — long-horizon investors betting on Intel's turnaround — sees the same numbers as the most encouraging part of the release, not the most alarming. Intel Foundry, the division that manufactures chips for outside customers, lost $2.1 billion in the quarter. That's real money, but it's narrower than the $3.2 billion loss a year earlier, even as foundry revenue grew 31%^[12]. Fabs carry enormous fixed costs, so a shrinking loss on rising revenue looks like a business filling up, not one falling behind. Intel also pulled forward its target for high-volume production of its next chipmaking process, called 14A, to 2028 — a year earlier than previously planned, and close to rival TSMC's own timeline^[11]. Getting there requires ordering expensive tools years in advance; Intel says tooling spending alone is up 40% from 2025^[5]. To this camp, a company that's beaten its own guidance seven quarters running and is now spending to hit a competitive deadline looks different from one spending to chase a trend^[5].

The Government Is Both Referee and Player

There's a fourth force shaping how much of that $20-billion-plus goes where: Washington. Congress raised the federal tax credit for U.S. chip-manufacturing investment from 25% to 35% of qualified spending, for facilities placed in service after December 31, 2025, under a law signed in July 2025^[16]. That's a meaningfully better deal than the 25% figure often cited in older coverage — more than a third of every qualifying dollar Intel spends on U.S. fabs effectively comes back. Intel's own remarks stress that its 2027 spending increase will happen "predominantly" inside the United States^[5][7], which lines up with taking full advantage of that credit.

There's also a deadline attached: one CHIPS-tracking publication reports that construction has to begin by the end of 2026 to qualify^[20]. If that's right, it gives Intel a plain, non-demand-related reason to move its spending earlier rather than later — a capex raise driven partly by a tax deadline isn't the same signal as one driven purely by customer orders, even though both show up as the same number in a press release.

And the government isn't just setting the rules here. It's also a shareholder, holding roughly 10% of Intel through the August 2025 agreement^[17][18]. That's the same equity stake sitting behind the $12,529 million charge that made Intel's quarter look like a loss on paper^[3][4]. When Washington negotiated its way into Intel's ownership structure, it also wired Intel's rising stock price directly into Intel's own quarterly numbers.

Same Facts, Different Headlines

How this quarter got covered split roughly along where each outlet's audience sits. U.S. business press mostly played it straight — CNBC listed the beat and the capex raise as separate facts without drawing a line between them, leaving readers to connect the dots on why the stock still fell^[2]. Yahoo Finance's initial headline, "Intel Stock Soars After Massive Q2 Beat," captured only the after-hours pop and skipped the reversal that followed — while a second Yahoo piece the same week called the capex plan "a quiet positive for the foundry business," showing how much the same facts bend depending on which day you're describing^[8][22].

Asian trade press read the story almost entirely through the manufacturing race rather than the stock price. Taiwan-based TrendForce led with Intel closing the gap to TSMC's roadmap; a spending increase, in that frame, is simply the cost of staying competitive, not a red flag^[11]. A Forbes contributor column, published before this earnings report, had already set up a bearish "AI spending fears" frame that the capex raise then slotted neatly into^[14] — a reminder that some of the story's shape was written before the numbers even came out.

What Isn't Settled Yet

Strip away the framing and the material facts are these: Intel sold more, in more places, than it did a year ago, and its foundry business is losing less money while growing faster^[1][6][12]. The $11 billion loss is a paper effect of Intel's own stock going up, not a sign of the business struggling^[3][4]. None of that resolves the actual question on the table.

The 14A process Intel is racing to ready doesn't reach high-volume production until a targeted 2028^[11]. No amount of favorable framing moves that date. Whether the spending was the right call will be decided by two things that don't exist yet in either direction: real chips coming off real production lines at usable yields, and signed contracts from outside customers willing to bet their own products on Intel's fabs.