Google introduced Gemini 3 — a top-tier AI model that outperforms major rivals on many benchmarks. Crucially, Gemini 3 was trained entirely on Google’s in-house TPUs, not Nvidia GPUs. Google is now offering TPUs to external customers as a lower-cost, integrated alternative to GPUs. Its full-stack offering (custom chips, software and cloud services) can reduce hardware and operational costs for large AI workloads.
C
CNBC TV1805-02-2026, 07:49

Nvidia, Broadcom Shares Soar After Alphabet's AI Capex Boost

  • Broadcom and Nvidia shares surged in extended trading following Alphabet's announcement of increased AI capital expenditure.
  • Alphabet plans to nearly double its capital expenditure to up to $185 billion this year, primarily for expanding data center capacity for AI workloads.
  • Melius Research's Ben Reitzes highlighted the significant benefit for Broadcom and other companies associated with Alphabet due to this spend.
  • Broadcom collaborates with Google on the design and manufacturing of its in-house TPUs and is expanding its custom chip business (ASICs) for AI.
  • Despite Google's proprietary TPUs, Alphabet continues to utilize Nvidia chips, contributing to Nvidia's stock rise.

More like this

Loading more articles...