The Intel foundry story just crossed the line from promise to proof. A major cloud provider has committed to build chips on the 18A node, the first outside customer the company can actually point to. Now Lip-Bu Tan has to keep delivering it, quarter after quarter, for a stock that already assumes he will.
By Blumefield | July 23, 2026
Intel handed Wall Street the revenue and earnings it asked for on Thursday, yet the real verdict on the Intel foundry turnaround was written days before the numbers landed. The company reported second-quarter revenue of $14.4 billion and non-GAAP earnings of $0.22 a share, both broadly in line with consensus. Gross margin arrived just below management's 39 percent guide. On any ordinary reading, that is a clean and unremarkable quarter. Intel in 2026 is not an ordinary story.
The shares have roughly tripled over the past year, one of the best runs of any large-cap stock, and they have done it on a single idea: that Intel can become the credible American alternative to Taiwan's chipmaking dominance. Every line of this report was read through that lens. Data Center growth, gross margin, and above all any hard evidence of outside demand for its factories mattered far more than the headline print.
The Number That Mattered Was Not the One Everyone Watched
For most companies, a quarter lives or dies on revenue and profit. For Intel right now, it lives or dies on proof of a business that barely exists yet. In the first quarter, Intel Foundry booked $5.4 billion in revenue, but nearly all of that came from Intel's own product divisions paying Intel's own fabs. The question investors have hammered on for two years is blunt. Will anyone else pay to have their chips made here?
This week the answer finally came with a name and a node attached. Intel confirmed that its 18A process had secured a manufacturing commitment from a major cloud service provider, the first externally recognized commercial customer at volume. Separately, the company said Fortinet will build its next-generation security processor on the Intel 4 node, the first named external foundry customer Intel has disclosed since Lip-Bu Tan took the top job. Two years of aspiration, condensed into two signatures.
Neither deal alone rebuilds a company that spent a decade losing its manufacturing lead. Together they change the conversation. An Intel foundry pitch stops being a slide the moment a buyer's logo is on it, and the market noticed. Shares jumped 8.4 percent on Tuesday as the trio of announcements landed, a rally that had little to do with the earnings that followed.
From Slideware to Signed Orders
The turnaround has always rested on one man's credibility. Tan became chief executive in March 2025 after Pat Gelsinger's exit, inheriting a company whose foundry ambitions were long on capital spending and short on named customers. His job was never to out-engineer Nvidia in AI accelerators. It was to make the manufacturing plan believable to people who write nine-figure cheques.
That list is no longer hypothetical. Beyond the cloud commitment and the Fortinet win, both Apple and Microsoft have been confirmed as 18A design partners, and Amazon Web Services previously selected Intel's factories for a custom AI fabric chip on 18A and a Xeon 6 part on the older Intel 3 node. The company has also been steadily upgrading its data center hardware, with select Xeon 6700P processors set to support faster 8000 MT/s memory when they ship in the coming weeks, a move aimed squarely at the agentic AI workloads now pulling attention back toward the CPU. Intel's official foundry roadmap now reads less like a wish and more like a schedule.
The strategic logic is cleaner than the AI-chip horse race the market obsesses over. Nvidia still owns the center of the accelerator market and is unlikely to be dislodged. Intel does not need to win there. The Intel foundry needs to be the most trusted advanced-node manufacturer on American soil at the exact moment Washington and hyperscalers are desperate for chipmaking capacity that does not run through Taiwan. That is a smaller ambition, and a far more achievable one.
The 18A Yield Curve Bends Toward Credibility
None of this works without yields, the unglamorous percentage of chips that come off a wafer functional rather than scrapped. Here the numbers moved the most. Intel foundry yields on 18A have climbed to roughly 85 percent, up from about 65 percent a quarter earlier. TSMC's leading N2 node sits near 90 percent, which means the gap that once defined Intel's decline has narrowed to something like a rounding error in strategic terms.
There was a second, quieter milestone. Equipment maker ASML confirmed on its own July earnings call that Intel is the first company to run high-volume logic production using High-NA extreme ultraviolet lithography, the most advanced patterning technology in the industry and a tool TSMC has been slower to deploy at scale. For a company whose entire crisis began when it fell behind on process technology, being first on the newest machine is more than symbolic. Intel's investor materials frame it as evidence that the manufacturing organization has genuinely turned. ASML's confirmation gave the claim an independent witness.
Yields at 85 percent do not just improve the economics of Intel's own products. They are the precondition for every external order the foundry hopes to win, because no serious customer commits volume to a process that scraps a third of its output. The Fortinet and cloud deals are believable precisely because the yield curve finally supports them, and Fortinet's willingness to put a shipping product on an Intel node is the kind of validation no press release can manufacture.
A Stock That Now Expects Everything
Here is the uncomfortable half of the story. Intel shares peaked above $142 in June before pulling back roughly 32 percent, dragged down as the broader Philadelphia semiconductor index slid into a bear market, off more than 20 percent from its high. Even after that retreat, the stock is up around 163 percent for the year, which by any traditional valuation measure leaves it expensive. Options markets were pricing a swing of about 12 percent in either direction around the print.
That is the real risk now, and it has almost nothing to do with whether Intel is improving. It clearly is. The danger is that a stock priced for a flawless execution has almost no room to absorb disappointment. A soft data center figure, a weaker-than-guided margin, or foundry commentary that lacks a dollar value could send it sharply lower even on a quarter that beats. Good news that is merely good, rather than spectacular, can still trigger a sell-off when expectations have run this far ahead.
The company is also still cutting, with fresh layoffs reported in its Data Center group as Tan pushes for focus and efficiency. A business can be genuinely healing and under real pressure at the same time. Both things are true here, and investors chasing the rally should hold them together rather than pick the one that flatters their position. For deeper market analysis, readers can follow Blumefield as the story develops.
What Tan Still Has to Prove
The honest read on Thursday is that Intel cleared the bar it was set without vaulting over it. Revenue and earnings met expectations, yields impressed, and the foundry finally has outside customers with names rather than placeholders. What it does not yet have is a large, disclosed stream of external foundry revenue with a dollar figure and a timeline, the metric that would let Wall Street re-rate the whole company as a growth story rather than a cyclical one.
That is the test for the coming quarters. Design wins from Apple and Microsoft need to convert into recognized revenue. The single cloud commitment needs to become several. The 85 percent yield figure needs to translate into contracts that management is willing to quantify out loud. Until then, the Intel foundry revival remains the most convincing comeback in American technology and, simultaneously, one of its most demanding stocks to own. The comeback is no longer the hard part. Proving it on a schedule, to an audience that has already paid for the ending, is. Blumefield will continue tracking the numbers as they land on Blumefield.