Google Adopts Nvidia Playbook to Scale TPU Business and Rival Chip Dominance

Google is expanding its custom TPU ecosystem by mirroring Nvidia’s strategy of financially backing AI data centers to drive customer adoption. This follows an order for 3 million TPUs from Intel for 2028 delivery and talks with Samsung to manufacture the memory I/O die for the Icefish TPU on 2 nm, while TSMC handles the 1.4 nm compute engine. The moves diversify supply, support Alphabet’s $460 billion backlog, and position newer TPU 8t and 8i chips to deliver 2-4x faster performance at 30% lower cost than prior generations.