A Second Source: Google Turns to Samsung for Its AI Chips

Ionic Global Research on 16 Jun 2026
sparklesAI Summary
Google is reportedly in talks to have Samsung manufacture part of its next-generation TPU on Samsung's 2-nanometre process, a notable break from its all-TSMC past. The trigger is scarcity: TSMC is at capacity, pushing Big Tech to cultivate a credible second source. It is still early, with talks rather than a signed deal and mass production not targeted until around 2028.
A Second Source: Google Turns to Samsung for Its AI Chips
A Second Source: Google Turns to Samsung for Its AI Chips. TSMC is maxed out. Big Tech wants a backup.

Google is breaking its all-Taiwan habit. Google designs its own AI chips — called TPUs — to lean less on Nvidia, and Taiwan's TSMC has always built everyone of them. This week, reports said Google is in talks to have Samsung manufacture part of its next-generation TPU (codenamed "Icefish") on Samsung's advanced 2-nanometre process — a notable break from its TSMC-only past.

image

Why now: TSMC is simply maxed out. The trigger is scarcity. TSMC's chairman publicly acknowledged on June 9 that customer demand is so high there are limits to handling it. So, Google is splitting the job: TSMC would make the main compute "brain" on its most advanced process, while Samsung would make a supporting piece — the part that links the chip to its memory.

It signals the chip making race is opening up. For years, TSMC has been almost the only option for cutting-edge chips. This is a sign Big Tech wants a second supplier — and a potential turning point for Samsung's long-struggling factory arm, which has recently also won orders from Tesla and Nvidia, pitching a one-stop package of chips, memory and packaging. Samsung already supplies more than 60% of the fast memory used in Google's current TPUs.

Why it matters for investors. A more competitive chip making market is healthy: it eases the AI bottleneck and creates a credible second source. It's also a reminder that the AI build-out lifts more than one boat — even with TSMC still dominant, rivals can win meaningful work. The caveat: this is still early — talks rather than a signed deal, with mass production not targeted until around 2028.

Ionic View

The wider story is that the AI chip supply chain is slowly broadening. TSMC remains the clear leader by a wide margin, but its capacity limits are pushing the biggest tech firms to cultivate a credible second source — and Samsung, after years of trailing, is rebuilding its standing with work from Google, Tesla and Nvidia. The direction of travel matters more than any single contract: a more contested chip making market spreads the AI-hardware opportunity a cross more players and lowers the industry's reliance on one supplier. We see that broadening as the real signal here.

Ionic Wealth Newsletter

Investment insights
made
simple

Sign up for our newsletter about wealth, markets, and more.