Google has placed an order for three million tensor processing units (TPUs) with Intel Foundry — one of the largest chip manufacturing contracts ever announced — as the AI giant works to secure the semiconductor supply it needs to power its rapidly expanding AI infrastructure. Simultaneously, Nvidia is reportedly exploring its own foundry partnership with Intel, a development that would mark a dramatic shift in the semiconductor industry’s supply chain dynamics and potentially transform Intel’s position in the AI chip ecosystem.
Google’s decision to turn to Intel for TPU manufacturing reflects both the strategic importance of diversifying chip supply beyond TSMC and growing confidence in Intel’s manufacturing capabilities under CEO Lip-Bu Tan. The order comes after Intel’s 18A process node entered risk production on schedule — a milestone that demonstrates Intel is executing on its foundry roadmap with the reliability that major customers require before committing billions of dollars in manufacturing commitments.
The three million unit order is enormous in scale, reflecting Google’s extraordinary appetite for AI computing infrastructure. Google’s TPUs power its AI services including Google Search AI features, Google Cloud AI infrastructure, Gemini model training and inference, and internal research. The company has been a consistent pioneer in custom AI silicon, developing its own chips rather than relying entirely on third-party GPUs — a strategy that has delivered both performance advantages and supply chain resilience.
Nvidia’s interest in an Intel Foundry partnership is perhaps the more surprising development. Nvidia has historically manufactured its GPUs exclusively through TSMC, but the chronic shortage of TSMC’s most advanced nodes — exacerbated by enormous demand for AI chips — may be motivating Nvidia to explore alternatives. Intel’s 18A process node, with its RibbonFET and PowerVia technologies, offers competitive density and performance characteristics that could make it viable for certain Nvidia product lines.
Intel has pledged over $100 billion toward global manufacturing expansion, with a focus on advanced nodes that can serve both internal product lines and external foundry customers. The combination of Google’s TPU order and potential Nvidia partnership would represent a major validation of Intel’s foundry strategy, potentially attracting additional major customers who have been watching the company’s progress before committing.
For the AI industry, the emergence of Intel as a significant AI chip manufacturer would have positive implications for supply security and pricing. A more competitive foundry market with three viable advanced-node manufacturers — TSMC, Samsung, and Intel — would give AI companies greater flexibility in sourcing the chips they need, reducing the concentration risk that has contributed to semiconductor shortages and elevated prices. The long-term health of the AI ecosystem depends on robust, resilient chip manufacturing capacity, and Intel’s resurgence as a foundry player contributes meaningfully to that goal.
