Intel Raised Capex and the Stock Went Up: What That Says About the AI Trade

Alphabet raised capital expenditure guidance Wednesday night and fell roughly 6 percent. Intel raised capital expenditure guidance Thursday night and rose.

That is the day's most useful data point, and it settles a question worth being precise about: the market is not punishing capital spending. It is punishing capital spending that cannot yet be tied to committed demand.

The Numbers

Intel reported second-quarter revenue of $16.1 billion against a $14.42 billion consensus — up 25 percent year over year, the fastest revenue growth for any quarter since the third quarter of 2011. Adjusted earnings of 42 cents per share roughly doubled the 21-cent estimate, swinging from a 10-cent adjusted loss a year earlier.

Guidance came in above consensus on both lines: third-quarter revenue of $15.8 to $16.8 billion against $15.1 billion expected, and adjusted earnings of 38 cents against 27 cents.

The segment detail is where the AI exposure shows. Data Center and AI revenue climbed 59 percent year over year to $6.3 billion. Intel Foundry revenue rose 31 percent to $5.8 billion. Client Computing gained 13 percent to $8.9 billion — meaning the manufacturing lines are running hot across the business rather than one strong segment masking weakness elsewhere. GAAP gross margin improved to 40.4 percent from 27.5 percent, and GAAP operating margin swung to positive 11.1 percent from negative 24.7 percent.

The $11 Billion Loss That Is Not a Loss

The headline that will confuse people: Intel reported a GAAP net loss of $11 billion, or $2.16 per diluted share, in a quarter where it beat on every operating metric.

The entire gap traces to a single non-cash line — a $12.5 billion mark-to-market loss on escrowed shares tied to Intel's CHIPS Act agreement with the US government. Intel holds shares in escrow for the Department of Commerce under that arrangement, and accounting rules treat the obligation as a derivative liability. As those escrowed shares gain value, Intel must book the increase as a loss.

Which produces a genuinely strange result: Intel reported an $11 billion loss substantially because its own stock became more valuable. The shares are up more than 170 percent this year after rising 84 percent last year, and that appreciation is what generated the charge. Strip the item out and non-GAAP net income was $2.2 billion.

This is worth understanding rather than dismissing, because it cuts both ways. It is not an operational problem, and treating it as one misreads the quarter entirely. But it is also not nothing — it is a real obligation whose accounting value moves with the share price, and it will reverse into a gain if the stock declines. An investor should know the mechanism rather than simply being told to ignore the GAAP line.

The Part That Answers This Morning's Question

Here is the detail that makes this more than a good quarter.

On the earnings call, Intel raised its 2026 capital expenditure forecast to more than $20 billion, a significant increase from earlier plans, with tooling spend alone up 40 percent from 2025. Management further flagged that 2027 capex will be significantly above 2026 levels. Chief financial officer David Zinsner described the company aggressively locking in tool purchase orders and accelerating clean room build-outs.

That is, structurally, the same announcement that took Alphabet down six percent and Taiwan Semiconductor down after a record quarter: we are spending more, and next year we will spend more still.

Intel rose.

The difference is in one sentence from Zinsner, and it is the whole distinction: the investments are gated by concrete customer commitments. He paired the spending increase with the phrase we remain very disciplined.

The supporting evidence is not rhetorical. Intel disclosed that it has reached long-term agreements with server CPU customers — some with pricing locked in, others focused on volume. Zinsner said the company is supply constrained, with data center customers demanding more than it can currently produce.

Two Companies, Same Words, Opposite Reactions

Both Alphabet and Intel used the phrase supply constrained this week. Both raised capital expenditure. The market rewarded one and punished the other, and the reason is not sentiment.

Alphabet is buying capacity. Its capital expenditure funds infrastructure it will operate itself, monetized through cloud revenue that arrives over subsequent years. The evidence it offered was genuinely strong — cloud revenue up 82 percent, operating margin expanding from 20.7 percent to 35.6 percent, a $514 billion backlog. But free cash flow turned negative at $5.9 billion, because capex grew 101 percent while cloud revenue grew 82 percent. The spending is compounding faster than the return, and the return remains partly a forecast.

Intel is selling capacity. Its capital expenditure builds manufacturing that customers have contracted to buy from. The long-term agreements convert forward demand from a projection into an obligation. When a company says it is supply constrained while holding signed volume and pricing commitments, the capital spending has a visible destination — it exists to fill orders that already exist.

That is the variable. Not the size of the spend, not the industry, not whether the story involves AI. The question the market is actually asking is whether the capital has a buyer attached to it yet, and Intel could answer yes in a way Alphabet structurally cannot.

It also explains Taiwan Semiconductor's fade last week despite a record quarter: TSMC raised its own capex roughly 15 percent and flagged margin dilution from overseas fab ramps — spending into future capacity rather than demonstrating conversion of past spending. Same category as Alphabet, different industry.

The Cash Flow Complication

One number deserves attention rather than celebration, because the bull case here has a real qualification.

Intel generated $7.0 billion in operating cash flow in the quarter — and adjusted free cash flow was negative $8.4 billion, against negative $1.05 billion a year earlier. The capital spending is enormous relative to what the business currently produces in cash.

So Intel is in the same structural position as Alphabet on the cash flow statement: spending more than operations generate, funding the difference from the balance sheet. The difference is that Intel's spending is backed by contracted orders and supported by CHIPS Act investment tax credits, and its revenue is accelerating at the fastest rate in fifteen years.

That is a better position. It is not an exemption from the same arithmetic. If the AI demand cycle turns before those fabs are filled, the contracts become the only thing standing between the spending and a problem — which, notably, is precisely why long-term agreements with locked pricing are becoming common across the semiconductor and memory industries. Vendors are trying to preserve current pricing and market power in case the market turns. The contracts are themselves a hedge against the cycle, and they exist because the people closest to it consider a turn possible.

What This Means for Reading the Rest of the Season

The useful frame going forward is not bullish or bearish on AI. It is a question to ask of any company reporting capital expenditure growth: can it point to committed demand for what the spending produces?

Companies that can — through contracted volume, locked pricing, or a supply constraint against orders already placed — have been rewarded even while raising spending. Companies that can only point to a backlog, a margin trend, or a strategic conviction have been sold even while beating on every operating line.

That distinction is more durable than any single quarter, and it is a more precise tool than the narrative that markets have simply soured on AI. They have not. They have started demanding that the capital chain show its counterparty.

It also produces a specific thing to watch. If contracted demand is what the market is rewarding, then the value of those contracts depends entirely on the counterparties honoring them — and the counterparties are, in many cases, the same hyperscalers whose own free cash flow just went negative. A chain of commitments is only as sound as the balance sheets behind it.

The Discipline

Intel moved several percent in after-hours trading on this release, in the window when protective orders cannot act, because a stop is an instruction to transact when a price is touched and no price is being touched between the close and the open. Two megacaps gapped the night before on the same category of news. That is three overnight repricings in two sessions, all on guidance rather than results.

The defenses do not require predicting which way any of them break. Position sizing that assumes a gap rather than a fill, capping maximum capital per trade at capital / 20 so a hostile overnight window is survivable rather than terminal. Limits on concurrent positions, because names sharing a factor are not a diversified book regardless of how many tickers they occupy. Hard daily loss limits. Exits defined before the print rather than improvised during the reaction.

That is why the execution and risk layer is worth engineering rather than improvising. StaxInvesting runs it as Software — Not Signals, self-hosted with zero account access, executing on a member's own connected brokerage under rules they set and enforcing limits mechanically rather than depending on a trader to parse an earnings call in real time. In a 2026 retail volatility regime, that bounds what a single print can cost. It does not make the print predictable.

The Bottom Line

Intel posted its fastest revenue growth since 2011, beat consensus by roughly 12 percent on revenue and doubled it on adjusted earnings, guided above expectations, raised 2026 capital expenditure above $20 billion, said 2027 will be higher still — and the stock rose. Alphabet beat, raised capex, and fell. The distinguishing variable was not the amount of spending or the quality of the quarter. It was whether the capital had committed buyers attached.

The $11 billion GAAP loss is an accounting artifact of Intel's own share price appreciation and should not be read as operational deterioration. The negative $8.4 billion adjusted free cash flow should not be waved away — it is the same spending-outruns-cash-generation dynamic visible across the AI complex, differentiated here by contracted demand rather than eliminated by it.

The market has not turned against AI capital spending. It has started asking who is on the other side of it.

Past performance does not guarantee future results, and nothing here is financial advice or a recommendation to buy or sell any security or options contract. Companies are named to illustrate market structure and are not endorsed or criticized as investments. Consensus estimates, management commentary, and forward guidance are third-party statements subject to revision. GAAP and non-GAAP measures differ materially in this reporting period and both are described for clarity rather than to favor either presentation. Market data reflects reporting as of July 23, 2026, including after-hours trading which is subject to change before the next open. Options trading involves substantial risk of loss and is not suitable for all investors; stop orders do not execute when markets are closed and do not guarantee an execution price, and overnight gaps can produce losses materially larger than intended. No risk setting or automation prevents losses or guarantees a profitable outcome. StaxInvesting provides self-hosted trading software — not signals, financial advice, or a managed account — that runs on the member's own connected brokerage; StaxInvesting never accesses member funds, credentials, or trades.