Google’s next AI-chip order puts Intel Foundry’s comeback to the test, while NVIDIA trials loom in the background

Google has chosen Intel to manufacture more than 3,000,000 tensor processing units, or TPUs, according to a market report published by Yahoo Finance. TPUs are specialized chips designed for machine learning workloads, particularly neural-network training and inference used in large-scale AI systems. The report says production is targeted for 2028, placing the procurement several years ahead of launch and giving Intel Foundry a long runway to ramp capacity and yield.

The size and timing of the order matters for Intel because external foundry manufacturing is a different business than Intel’s traditional approach of building most chips it sells. Intel Foundry, the segment focused on contract manufacturing for other companies, has been working to restore momentum after years of production challenges and an industry-wide shift toward relying more on external and specialized manufacturing capacity for leading-edge chips.

A key takeaway from the report is that the Google TPUs are expected to be Intel’s largest external manufacturing commitment to date, according to the framing in the Yahoo Finance piece. If accurate, that would represent a major validation of Intel’s ability to handle complex, high-volume production requirements for an AI-focused customer at scale, not just smaller or experimental orders.

For Google, the decision reflects a balancing act that many AI leaders are now forced to manage: performance requirements, supply assurance, and manufacturing diversification. Using a mix of silicon platforms and manufacturing partners can reduce single-source risks and support long-term roadmap planning. The 2028 target also indicates that the deal is less about near-term delivery and more about securing future chip roadmaps and manufacturing capability well in advance.

For the broader AI hardware market, the report’s mention of “NVIDIA trials” alongside Intel’s win points to how tightly connected buyers, chip designs, and manufacturing are in this cycle. NVIDIA’s GPUs and software ecosystem dominate many AI data centers, but the supply chain is not static, and large customers evaluate multiple paths for compute. The Yahoo Finance report indicates that NVIDIA is part of the landscape under consideration, though it does not provide sufficient detail in the available material here to specify what the trials involve, what hardware is being tested, or whether they are linked to the Google order.

Intel’s situation is especially sensitive to whether it can sustain consistent production quality and cost as workloads scale. In advanced chip manufacturing, yield and defect rates can make or break economics, particularly for high-density accelerators where a small failure rate can translate into large losses at volume. A large TPU commitment would therefore test not only capacity planning but also process maturity for Intel’s foundry offerings.

Still, the publicly available detail for this story appears limited. The Yahoo Finance report describes the scale of the TPU selection and the 2028 production target, but the information available here does not clarify contract terms, which specific TPU generations or variants are involved, delivery schedules, or how performance and power requirements will be validated. It also does not provide specifics on what “NVIDIA trials” means in operational terms.

Looking ahead, market participants will likely focus on whether Intel Foundry can turn large customer commitments into measurable milestones, such as announced process targets, partner qualification timelines, and any later updates on production ramp. For Google, the next indicates will likely come through future product and infrastructure announcements that indicate where these TPUs are used. For NVIDIA and other AI hardware suppliers, the watch item is whether customers continue to diversify chip supply and testing efforts beyond the current mainstream deployments. Neither side’s longer-term economics can be inferred from this report alone, but the selection itself suggests a meaningful shift in who is considered for high-volume AI chip fabrication.