Sources: Google recently placed an order with Intel to manufacture 3M+ TPUs in 2028; Nvidia is testing Intel's tech for a new processor and running 18A trials
TSMC's capacity struggles are turning into a boon for Intel. — As the Taiwanese chip making giant struggles to meet overwhelming demand …
Context & Ripple Effects
Related coverage has shown Google expanding TPU ambitions beyond internal use, including pitching the chips to external customers, while its next-generation Icefish design is already being split across suppliers: Samsung for a memory I/O die and TSMC for the compute engine.
At the same time, demand for Nvidia hardware and advanced packaging has put pressure on TSMC’s supply chain. Earlier reports that Nvidia, Broadcom, and AMD were evaluating Intel’s 18A process make the reported Google order and Nvidia trials a more consequential test of Intel’s foundry strategy.
First-order effects
- Intel would gain a large prospective 2028 TPU manufacturing customer in Google, providing a concrete demand anchor for its advanced-node foundry capacity.
- Nvidia’s reported 18A trials and processor technology testing give Intel a second high-profile potential customer, while Google gains another manufacturing option for TPU supply.
Second-order effects
- TSMC faces a clearer risk that major AI-chip customers will distribute future production across alternative foundries when leading-edge wafer or packaging capacity is constrained.
- Google’s reported use of Intel alongside planned Samsung and TSMC roles reinforces a multi-supplier model for AI accelerators, increasing the importance of coordinating chip design, memory I/O, compute dies, and packaging across vendors.
Third-order effects
- If these evaluations turn into sustained production, Intel could shift from being principally a prospective alternative to TSMC into a meaningful second source for leading AI silicon, improving buyers’ leverage and supply resilience.
- The broader AI-chip supply chain may become less centered on a single foundry relationship and more modular across fabrication and packaging partners, though Intel still must convert trials and reported orders into dependable high-volume output.
The trend: AI-chip buyers are diversifying advanced manufacturing and packaging supply as accelerator demand outpaces the capacity of the incumbent supply chain.