• Google (GOOGL) and NVIDIA are exploring Intel as an additional or backup manufacturing partner, focusing on its advanced packaging and process nodes like 14A and 18A.
  • The move reflects a broader industry push to diversify supply chains amid surging AI chip demand and geopolitical risks.
  • Intel's foundry ambitions could get a major boost, though near-term feasibility hinges on yield, cost, and capacity.

Strategic Diversification

Google and NVIDIA are in early discussions with Intel to use its manufacturing capabilities as a backup for producing AI accelerators and data-center chips, according to people familiar with the matter. The talks center on Intel's advanced packaging technology (EMIB) and its upcoming 14A and 18A process nodes, which could serve as alternatives to TSMC's dominant foundry services.

“We are always evaluating multiple suppliers to ensure resilience and performance,” a Google spokesperson said, declining to comment on specific partnerships. NVIDIA did not respond to requests for comment. Intel declined to comment on customer discussions.

The potential partnerships come as hyperscalers and chip designers race to secure capacity for AI workloads. Google custom designs its Tensor Processing Units (TPUs), while NVIDIA’s accelerators dominate the market. Both rely heavily on TSMC, creating concentration risk that industry executives have flagged.

“Having a second source for advanced manufacturing is prudent,” said a semiconductor analyst who requested anonymity. “Intel’s foundry is unproven at scale for cutting-edge AI chips, but its packaging and U.S. location are compelling.”

Intel’s Foundry Push

Intel has invested heavily in its foundry business under CEO Pat Gelsinger, aiming to compete with TSMC and Samsung. The company is building new fabs in Ohio and Arizona, and its 18A node is slated for production by 2025. However, Intel has yet to land a large-scale customer for its most advanced processes.

“Intel’s process technology is competitive, but the company must prove it can deliver on time and at high yields,” said a former Intel executive. For NVIDIA and Google, even limited use of Intel for packaging or specific chip components could provide a strategic hedge.

The discussions also align with U.S. policy goals. The CHIPS Act allocates $53 billion to boost domestic semiconductor manufacturing, and Intel is a primary beneficiary. Dual-sourcing strategies could help mitigate geopolitical risks tied to Taiwan, analysts note.

Implications

If finalized, the partnerships would reshape the foundry landscape, legitimizing Intel’s comeback in manufacturing. For Google and NVIDIA, it provides pricing leverage and supply security. But obstacles remain: Intel’s capacity is still ramping, and its technology must match TSMC’s maturity.

“The base case is that Intel becomes a meaningful second source for edge computing and certain data-center workloads, not a full replacement,” said the analyst. Any deal would likely start with low-volume pilots.

Correction: An earlier version of this article incorrectly stated that Intel’s 18A node is in production. It is scheduled for 2025.