Google Cloud Challenges Nvidia with Its New TPU 8 AI Chips

Google Cloud Unveils Its New AI Chips TPU 8

Google Cloud recently announced the launch of its eighth generation of artificial intelligence chips, known as Tensor Processing Units (TPUs). These new chips, unveiled on Wednesday, come in two distinct versions: the TPU 8t, dedicated to training models, and the TPU 8i, designed for inference. In this context, inference refers to the ongoing use of models after users have submitted their queries.

Google's new TPUs boast impressive performance specifications. They offer up to three times faster training for AI models and an 80% improvement in performance per dollar compared to previous generations. Additionally, these chips allow for more than one million TPUs to operate together in a single cluster, promising increased computing power at a reduced energy and financial cost.

Google refers to these chips as TPUs, not GPUs, because they were initially named Tensor due to their low energy consumption.

A Complementary Strategy to Nvidia

While these technological advancements are significant, they do not represent a direct attack on Nvidia, at least for now. Google, like other major cloud providers such as Microsoft and Amazon, is using these new chips to complement the Nvidia-based systems already present in its infrastructure. Google also plans to integrate Nvidia's latest chip, Vera Rubin, into its cloud later this year.

The collaboration between Google and Nvidia does not stop there. The two companies are working together to enhance the efficiency of Nvidia-based systems in Google's cloud. This includes strengthening the software-defined networking technology Falcon, developed by Google and open-sourced in 2023 under the auspices of the open-source data center hardware organization, the Open Compute Project.

An Uncertain Future for Nvidia?

Although hyperscalers like Google, Amazon, and Microsoft are developing their own AI chips, Nvidia remains a major player in the market. Nvidia's market capitalization is nearing $5 trillion, a testament to its dominant position. According to analyst Patrick Moore, while the rise of Google's TPUs may seem threatening, Nvidia continues to thrive.

Ultimately, Google's growth as an AI cloud provider could even benefit Nvidia by increasing demand for its chips, despite the growing use of Google's TPUs.