TPU Tech Tussle: Google Cloud Targets AI Chip Turf
Google Cloud has unveiled two powerful eighth-generation chips of its own design—TPU 8t and TPU 8i—in a bid to expand its dominance in the artificial intelligence (AI) technology market. These chips are positioned as alternatives to offerings from Nvidia, which currently holds a dominant position in the AI chip sector.
Google stated that the two new chips are designed for different purposes. The TPU 8t is specifically engineered for large-scale AI model development, or “training.”
On the other hand, the TPU 8i is built for “inference,” enabling faster generation of responses or outputs from AI models. Compared to the previous generation, these new versions are expected to deliver AI training speeds up to three times faster while reducing customer costs by nearly 80 percent.
According to market analysts, despite launching its own chips, Google is not immediately seeking to displace Nvidia from the market. Rather, like Microsoft and Amazon, Google aims to reduce its dependence on Nvidia within its cloud ecosystem and strengthen alternative solutions. The company also announced that Nvidia’s latest Vera Rubin chip will be made available to Google Cloud users later this year.
Currently valued at close to $5 trillion, Nvidia remains unrivaled in the AI chip market. However, the growing competition from tech giants such as Google, Amazon, and Microsoft in developing their own chips could reshape the market landscape in the long term. For now, Nvidia and Google have also announced plans to collaborate on advancing networking technologies as partners.
DBTech/BMT/OR
