According to Google, its new AI supercomputer is faster and more efficient than its competitors. Nvidia is the dominant company for AI model training and deployment, making over 90% of Tensor Processing Units (TPUs). AI models need many computers and hundreds to thousands of chips to be trained. Google announced it had built a system with over 4,000 TPUs that is faster and more efficient than the Nvidia A100.
Google has been falling behind slightly in AI, catching up with Microsoft’s Bing release as well as with Nvidia’s TPU production. This new development could indicate that Google is getting ahead, though their TPU results were not as impressive compared with the latest Nvidia AI chip.
Read more at CNBC.
Comments