Intel just landed the validation it desperately needed. Google has committed to ordering more than 3 million of its custom Tensor Processing Units (TPUs) from Intel for 2028 production, while Nvidia is actively evaluating Intel's advanced manufacturing capabilities for future chip designs. The announcements, which arrived within a 72-hour window, sent Intel stock soaring 12.45 percent to $111.79 and represent a fundamental reshuffling of how the world's biggest AI companies source their chips.

For years, there was essentially one answer to the question "Where do you manufacture cutting-edge AI chips?" Taiwan Semiconductor Manufacturing Company (TSMC). But that era is ending, not because TSMC is failing, but because demand for advanced chip production has simply outpaced what even the world's best foundry can deliver. Intel's emergence as a credible alternative isn't just good news for the company; it's a structural shift in how the AI industry manages risk.

Why Is Intel Suddenly in the Game?

The answer lies in a bottleneck that has quietly become the AI industry's most pressing constraint. TSMC's advanced packaging capacity, particularly its CoWoS (Chip-on-Wafer-on-Substrate) technology, is the critical chokepoint. CoWoS is the specialized process that connects high-bandwidth memory to logic chips, and it's essential for every advanced AI accelerator on the market. Demand has exploded so dramatically that TSMC literally cannot keep up.

This capacity crisis created an opening Intel has been waiting for. Under CEO Lip-Bu Tan, the company has spent the past year courting major tech customers with a simple pitch: we can be your second source. Google and Nvidia just said yes.

Google's decision is the anchor of this story. The company is already selling TPU compute capacity to giants like Apple and Meta, making TPUs central to Google Cloud's revenue growth. By splitting production between TSMC and Intel, Google insulates itself from a catastrophic supply disruption. Morgan Stanley estimates Google could produce over 6 million TPUs across 2027 and 2028, meaning Intel could handle almost half that volume.

What Exactly Are Google and Nvidia Testing?

The two companies are taking markedly different approaches to their Intel experiments, and understanding the distinction matters for what comes next.

Google has already signed a formal contract. The company will use Intel's 18A process node, which represents Intel's most advanced manufacturing technology to date. The 18A node uses two key innovations: RibbonFET, Intel's gate-all-around transistor architecture, and PowerVia, a backside power delivery system. Both deliver real density and performance improvements over Intel's previous generation.

The initial order covers 3 million units for 2028, with potential expansion to 6 million units across 2027 and 2028 depending on whether Intel hits yield milestones. The estimated revenue contribution ranges from $4 billion to $7 billion for the initial order alone, with the full expansion case potentially reaching $14 billion.

Nvidia's situation is different. The company is running early trials on Intel's 18A process, including multiproject wafer runs, to test whether Intel can manufacture a processor combining four graphics chips into a single unit for Nvidia's upcoming "Feynman" GPU architecture scheduled for 2028 release. Critically, Nvidia has not placed a firm order yet. For Nvidia, this evaluation is partly about immediate manufacturing needs and partly about leverage. If Intel proves itself, Nvidia suddenly has a strong negotiating position with TSMC, which currently holds all the cards.

How Does Intel's Technology Actually Compare?

The technical specs on Intel's 18A node are competitive with TSMC's N2 (2-nanometer) process at the design level. TSMC's N2 is currently in pilot production with high-volume manufacturing targeted for late 2026, while Samsung's competing 2-nanometer process has struggled with yield challenges.

But here's what matters most: the question that drove Google's decision wasn't whether Intel's chips looked good in a lab. It was whether Intel could execute on yields at the scale required for millions of units. The Google contract structure suggests that Intel's yield trajectory has hit the milestones the customer needed to see under nondisclosure agreements. That execution proof is now visible to every other fabless customer evaluating the Intel option.

Beyond the chip-making itself, Google is also looking closely at Intel's EMIB (Embedded Multi-die Interconnect Bridge) technology, which is Intel's direct competitor to TSMC's CoWoS packaging approach. EMIB functions as a high-speed data shuttle system between different chiplets inside a processor. MediaTek has already signed on to use Intel's EMIB-T for Google's next-generation TPU custom AI chips, and SK Hynix is testing compatibility of its high-bandwidth memory modules with Intel's packaging technology.

Steps to Understanding Intel's Foundry Strategy

• The Google Order: A formal 3 million unit commitment for 2028 production on the 18A process, with potential expansion to 6 million units, representing $4 billion to $14 billion in revenue depending on execution and volume.
• The Hitachi Partnership: A five-year strategic collaboration announced June 5 covering foundry tooling, quantum computing, energy optimization, custom silicon for edge AI, and factory automation, embedding Hitachi into Intel's manufacturing stack across multiple dimensions.
• The Nvidia Evaluation: Active testing of Intel's 18A process and advanced packaging for Nvidia's Feynman GPU architecture, with a decision window pointing to the third quarter, though no firm order has been placed.
• The Packaging Advantage: Intel's EMIB technology offers an alternative to TSMC's CoWoS, with early adoption from MediaTek and testing from SK Hynix, potentially giving Intel a competitive edge even if chip-making execution remains uncertain.

What About Intel's Track Record?

This is where skepticism is warranted. Intel has a long history of over-promising on process nodes and under-delivering on yields. The foundry division remains unprofitable, burning roughly $5 billion per quarter to build out capacity. Wall Street's consensus price target sits at $89, well below the current trading price of $111.79, and analyst ratings remain cautious with 31 hold ratings versus 12 buys and 5 sells.

The real test won't come until 2027, when production-scale yield data becomes visible. Until then, Google's order is a vote of confidence, but it's not a guarantee. The actual production-scale yield data will not be public until late 2026, but the Google contract structure suggests the data Intel has shared under NDA was strong enough to clear a major customer's qualification process.

Why Does Geopolitics Matter Here?

There's a political dimension to these deals that extends beyond pure supply chain optimization. The U.S. government has poured billions into bringing semiconductor manufacturing back to American soil through the CHIPS Act. Intel is arguably the biggest beneficiary of that push.

"Beyond the standard need to diversify, Google and Nvidia are even more motivated than usual to work with Intel. Supporting Intel supports U.S.-based manufacturing, which is important for the relationship with the U.S. administration," stated Gil Luria, analyst at D.A. Davidson.

Gil Luria, Analyst at D.A. Davidson

At a time when the Taiwan Strait remains a geopolitical flashpoint, having a stateside backup isn't just a nice business hedge. It's becoming a strategic necessity. The Trump administration has steered billions in incentives toward Intel, and Washington officials have actively encouraged other companies to use Intel's products.

What Does This Mean for the Broader AI Supply Chain?

The headline narrative is often framed as "Intel replacing TSMC," but that's not what's happening. TSMC remains the dominant force in advanced chip manufacturing. What's actually happening is resilience building.

By having a credible second source for manufacturing, the AI industry just got significantly safer. A single outage, a single geopolitical event shutting down a TSMC fab, would cripple the global AI economy. By bringing Intel into the fold, Google and Nvidia are building a moat around their own futures and reducing systemic risk across the entire industry.

Intel's stock has already responded, climbing 196 percent year to date, putting it in the top three performers across the S&P 500 for 2026. But the real test is whether Intel can deliver on the promise. Google's order suggests they believe it can. Whether that confidence is justified will become clear over the next 18 months as production ramps begin.