Intel's Fastest Revenue Growth in 15 Years Can't Mask a Foundry Problem — Stock Falls 8%

Intel had a number ready to silence every skeptic: 25% revenue growth in the second quarter of 2026, the company’s fastest expansion since 2011. Its data center business shot up 59%. Its advanced manufacturing process, 18A, reached yields of 85% — up from 65% the prior quarter. CEO Lip-Bu Tan called it proof that “AI is driving unprecedented demand for compute.” For a single afternoon, it looked like the most improbable turnaround in semiconductor history was working.

Then came the fine print. Intel’s foundry business brought in roughly $5.8 billion in revenue for the quarter — but external customer revenue, the metric that actually validates the company’s ambition to become a contract manufacturer for the world, amounted to approximately $174 million. Investors did the math overnight and sold. Intel closed down 7.89% at $92.32 on July 24, one day after reporting results that by almost every other metric should have sent the stock higher.

The Numbers That Impressed

Start with the headline: Intel’s total Q2 2026 revenue topped $16 billion, beating analyst consensus. The Data Center and AI segment — the business unit that houses Intel’s Xeon server processors and accelerator products — grew 59% year-over-year to $6.3 billion. That kind of data center growth hasn’t been seen since the earliest days of the cloud computing transition.

The company’s client computing group, which makes chips for PCs and laptops, also held its own amid signs of an AI PC upgrade cycle gaining traction. Consumers and enterprise IT departments alike are beginning to replace aging hardware with systems that can run local AI workloads — Intel’s Core Ultra processors with integrated NPUs are designed precisely for that use case.

Most critically, Intel Foundry — the business unit tasked with manufacturing chips for outside customers — grew revenue 31% compared to a year ago. The 18A process node, once derided as vaporware by industry analysts, has now achieved commercial-grade yields. Apple and Microsoft have both confirmed participation as early design partners. Intel says it has signed ten long-term foundry agreements with unnamed customers, a significant jump from the handful that existed six months prior.

“We are now supply-constrained in our data center segment — a problem we have not had in years,” Tan told analysts on the earnings call. “This is a fundamentally different Intel than the one we were 18 months ago.”

The Number That Deflated Everything

The 18A process exists. The yields are real. The customers, however, are not yet showing up in the revenue line in any meaningful way.

The $174 million in external foundry revenue represents roughly 3% of Intel Foundry’s total quarterly haul. The rest comes from manufacturing chips for Intel’s own product divisions — a figure that looks better on paper than it is in practice, because those transfers are accounted for at internal rates rather than external market prices.

For Intel’s foundry ambitions to justify the tens of billions in capital expenditures the company has committed, it needs names like Amazon, Google, Qualcomm, or Broadcom to commit serious wafer volumes to the 18A node. So far, the public evidence for that happening is thin. Apple and Microsoft are design partners — but design partnership agreements, which allow customers to test a process node before committing to volume production, are explicitly not the same as manufacturing contracts.

Wall Street’s reaction on July 24 reflected precisely that anxiety. Semiconductor peers moved with Intel: AMD fell 3.29% to $521.95, and Texas Instruments dropped 1.90%. The message from investors was clear: an Intel comeback that runs entirely on internal demand is not the same as an Intel comeback that is reshaping the global foundry market.

Why the 18A Story Matters So Much

To understand why the external foundry question dominates every Intel earnings call, it helps to understand what Intel is actually attempting. The company is trying to become the first American chipmaker in decades to offer cutting-edge contract manufacturing — competing directly with Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung Foundry for the orders that power the world’s most advanced electronic devices.

TSMC currently manufactures chips for Apple, Nvidia, AMD, Qualcomm, and most other leading fabless semiconductor companies. It has an estimated 50-55% share of the global contract chip market and dominates the sub-5 nanometer advanced segment. Intel’s pitch to these same customers is straightforward: manufacture with us in the United States, reduce your geopolitical risk, and get access to Intel’s advanced packaging and manufacturing capabilities.

The pitch has found an audience in Washington, where policymakers have aggressively pushed for domestic semiconductor production through the CHIPS and Science Act. Intel has received billions in federal support. But political backing and government grants don’t automatically translate into customer orders. Fabless chip designers need to be convinced that Intel Foundry can deliver their designs at competitive yields, competitive pricing, and on schedule — three areas where TSMC has decades of refinement.

The 85% yield figure on 18A is the key data point that Intel needed to make this argument credible. For context, a yield above 80% is generally considered commercially viable for volume production. Six months ago, some industry analysts were questioning whether 18A yields would reach commercially attractive levels before late 2026 or even 2027. The Q2 report suggests Intel is ahead of that skeptical timeline.

The Case for Patience

Not everyone interpreted Thursday’s sell-off as a death knell for the Intel turnaround narrative. Several analysts issued notes on Friday morning arguing that the stock reaction was an overreaction driven by investors misunderstanding the foundry revenue timeline.

The argument goes roughly like this: external foundry customer agreements involve multi-year design cycles. A company that decides today to tape out a chip on Intel 18A will not place production wafer orders for 12 to 18 months. The ten long-term agreements Intel announced are functionally an order backlog — they represent future revenue that simply hasn’t materialized yet because the customers are still in the design phase.

Intel’s 163% year-to-date gain heading into the earnings report also meant the bar for a positive stock reaction was extremely high. The company needed not just good results but blow-out results plus clear evidence of external demand — a combination that no single earnings quarter was likely to deliver simultaneously.

CEO Lip-Bu Tan has repeatedly emphasized that the Intel foundry transformation is a multi-year project. He set an explicit target of generating $15 billion in annual external foundry revenue by 2030. Against that timeline, $174 million in one quarter is not a crisis — it is simply the early innings of a long build.

What Comes Next

The next major catalyst for Intel’s foundry story will likely come in October, when the company is expected to name additional 18A customers and provide updates on the production ramp timeline for existing design partners. Any announcement naming a major fabless customer — a Qualcomm, a MediaTek, or a hyperscaler doing custom silicon — would be a significant inflection point for the stock.

In the meantime, Intel’s core businesses are genuinely thriving. Data center demand for Xeon processors accelerated as enterprises built out the CPU-side infrastructure needed to support their AI deployments. The AI PC cycle is creating a real PC upgrade tailwind for the first time since the pandemic-era surge. And the Foundry segment, whatever its external revenue limitations, is at least no longer a source of embarrassing headlines about process delays and yield failures.

Intel’s Q2 2026 report is the story of a company that has done everything it needed to do to get back in the game — and is now waiting for the market to verify that its bets on the future were correctly placed. The verification process, it turns out, takes longer than a single earnings quarter.

For investors, the question is whether the 15-year record revenue growth is the beginning of a sustained structural recovery or a peak driven by cyclical AI demand that will prove harder to sustain. That answer won’t come from quarterly revenue figures. It will come from the customer logos on the next foundry announcement.