Back to Blog
April 22, 2026 • Julie Bort

Google Cloud Unveils New AI Chips to Challenge Nvidia

Google Cloud Unveils New AI Chips to Challenge Nvidia Google Cloud Unveils New AI Chips to Challenge Nvidia Google Cloud Unveils New AI Chips to Challenge Nvidia

Google Cloud Launches New AI Chips to Compete with Nvidia

Google Cloud has announced its eighth generation of custom-built AI chips, known as Tensor Processing Units (TPUs), which will be divided into two distinct lines: the TPU 8t for model training and the TPU 8i for inference.

Key Features and Performance Gains

  • Dual Chip Strategy: The new TPUs are split into training-focused (TPU 8t) and inference-focused (TPU 8i) models.
  • Performance Improvements: Google claims up to 3x faster AI model training, 80% better performance per dollar compared to previous generations, and the ability to scale to over 1 million TPUs in a single cluster.
  • Efficiency: The new chips are designed to offer more compute power with less energy consumption and reduced costs for customers.
  • Custom Design: Google refers to these as TPUs, emphasizing their custom, low-power design originally named Tensor.

Competitive Landscape and Nvidia Integration

  • Supplementing, Not Replacing: Google's strategy, similar to other major cloud providers like Microsoft and Amazon, is to use its custom chips to complement its existing Nvidia-based infrastructure rather than outright replace it.
  • Continued Nvidia Support: Google Cloud will continue to offer Nvidia's latest chips, including the Vera Rubin, later this year.
  • Partnership on Networking: Google is collaborating with Nvidia to enhance the efficiency of Nvidia-based systems within its cloud through improved networking technology, specifically by optimizing Falcon, an open-sourced networking technology developed by Google.
  • Market Dynamics: Despite the development of in-house AI chips by hyperscalers, Nvidia remains a dominant force in the market, with its market capitalization reflecting its strong position.

Strategic Implications

  • Hyperscaler Evolution: As hyperscalers like Google, Amazon, and Microsoft continue to develop their own AI chips, their reliance on Nvidia may decrease over time, especially as enterprises increasingly adopt cloud-based AI solutions.
  • Ecosystem Integration: Google's efforts to improve networking for Nvidia chips highlight a pragmatic approach to leveraging existing strengths while developing future capabilities.

Event Announcements

  • StrictlyVC San Francisco 2026: The article includes a promotion for the StrictlyVC event in San Francisco on April 30, 2026, emphasizing networking and insights from industry leaders.
  • TechCrunch Disrupt 2026: Another event promotion is for TechCrunch Disrupt 2026 in San Francisco from October 13-15, 2026, targeting founders, investors, and tech leaders for networking and innovation opportunities.