Google takes on NVIDIA with new AI chips
What's the story
Google Cloud has unveiled its eighth generation of custom-built artificial intelligence (AI) chips, or tensor processing units (TPUs). The new range is divided into two variants: the TPU 8t for model training and the TPU 8i for inference. This strategic move comes as part of Google's larger plan to strengthen its position in the AI hardware market and take on industry leader NVIDIA.
Performance boost
Major performance boost over previous generation
The new TPUs are said to provide a major performance boost over their predecessors. Google claims these chips can deliver up to three times faster AI model training, 80% better performance per dollar, and the ability to cluster over one million TPUs in a single cluster. This means users will get more computing power while using less energy and money than before.
Strategic partnership
Google not directly taking on NVIDIA
Despite the impressive specs of its new TPUs, Google isn't directly taking on NVIDIA. Instead, like other major cloud providers such as Microsoft and Amazon, it is using these chips to supplement the existing NVIDIA-based systems in its infrastructure. The company has also promised that its cloud will support NVIDIA's latest chip, Vera Rubin, later this year.
Tech upgrade
TPU 8i can cluster up to one million chips
The TPU 8i comes with a whopping 384MB of SRAM per chip, three times more than the previous Ironwood version. This upgrade allows for faster data access and lower latency, which is critical for real-time AI interactions. The infrastructure can scale aggressively too, with Google claiming it can cluster up to one million chips for the most demanding workloads.
Business strategy
Google consolidates AI offerings under 'Gemini Enterprise'
The chip overhaul is part of Google's larger push into enterprise AI, where it sees the most reliable revenue opportunities. At its annual cloud conference in Las Vegas, executives made it clear that AI agents are now at the heart of its strategy. Google is also consolidating its AI offerings under "Gemini Enterprise," expanding tools like Vertex AI to help businesses build and deploy custom AI agents.