TSMC Shortage Drives Google, NVIDIA to Intel

According to a report by The Information, several leading AI chip developers, including Google and NVIDIA, are increasingly considering Intel as an alternative manufacturing partner for advanced processors as Taiwan Semiconductor Manufacturing Company (TSMC) faces mounting capacity constraints driven by surging artificial intelligence demand.

The ongoing AI boom has placed unprecedented pressure on TSMC's manufacturing infrastructure. Industry sources indicate that both the company's cutting-edge wafer fabrication lines and advanced packaging facilities used to integrate processors with High Bandwidth Memory (HBM) are operating at full capacity, with production slots largely booked out years in advance.

TSMC Chairman and CEO C.C. Wei previously warned shareholders that global semiconductor supply may struggle to fully meet AI-driven demand over the coming years, highlighting the industry's ongoing capacity challenges.

Among potential customers, Google appears to be moving the fastest. According to sources familiar with the matter, the company has spent several months evaluating Intel's advanced packaging technologies and has recently placed orders that could result in the production of more than three million of its custom Tensor Processing Units (TPUs) in 2028. A recent forecast from Morgan Stanley estimates that Google could manufacture over six million TPU chips between 2027 and 2028.

NVIDIA has not yet committed to production orders, but sources indicate that the company is assessing whether Intel's technologies can support its next-generation Feynman GPU architecture, expected to launch around 2028 and designed to integrate four GPU chiplets into a single package. NVIDIA is also reportedly conducting early multi-project wafer (MPW) testing using Intel's most advanced 18A process technology, a critical step in evaluating future manufacturing compatibility.

Interest in Intel's advanced packaging ecosystem extends beyond AI processor developers. Memory industry leader SK hynix is reportedly testing the interoperability of its memory products with Intel's Embedded Multi-die Interconnect Bridge (EMIB) packaging technology, reflecting broader industry efforts to optimize high-performance AI computing platforms.

While the sharp increase in AI-related semiconductor demand has created a significant opportunity for Intel's foundry business, the company still faces important execution challenges. Intel must demonstrate its ability to overcome past manufacturing delays, convert customer evaluations into large-scale production commitments, and successfully ramp its 18A process technology to volume manufacturing. Achieving these milestones will be essential for Intel to establish itself as a competitive player in the advanced semiconductor foundry market and challenge TSMC's dominance in next-generation chip production.