Google Cloud launches two new AI chips to compete with Nvidia
TechCrunch • • business
Key Points:
- Google Cloud announced its eighth generation of custom AI chips, splitting them into TPU 8t for model training and TPU 8i for inference tasks, aiming to enhance AI processing efficiency.
- The new TPUs offer up to three times faster model training, 80% better performance per dollar, and can scale to over one million units in a single cluster, promising more compute power with lower energy and cost.
- Despite these advancements, Google continues to supplement its AI infrastructure with Nvidia GPUs and plans to offer Nvidia’s latest Vera Rubin chip later this year, rather than fully replacing Nvidia hardware.
- Industry analysts note that while cloud providers like Google, Amazon, and Microsoft are developing their own AI chips, Nvidia remains dominant with a nearly $5 trillion market cap, and increased AI cloud growth could benefit Nvidia’s business.
- The development reflects a gradual shift where hyperscalers may rely less on Nvidia over time, but currently, Nvidia’s technology remains central to AI cloud services.
