By Tony Pialis, CEO and cofounder, Alphawave Semi
Tony Pialis
The AI computing war will reach a fever pitch
The race for evermore powerful AI supercomputers will shift into high gear and the industry will see more records being broken and set in 2023. Microsoft and Nvidia just announced plans to build a new Azure-hosted AI supercomputer, Nvidia and IBM releases more powerful chips, and newcomers such as Cerebras raised $720M in funding along with unveiling a 13.5M core AI supercomputer. This AI supercomputer arms race shows no sign of stopping as ML training becomes more complex. One emerging challenge, however, is being able to feed data fast enough with higher bandwidth and low latency to operate AI workloads to their full potential. This will drive the development of more highly customized and optimized silicon solutions with faster chip-to-chip connectivity links in the most advanced process nodes designed in leading-edge foundries like TSMC. The rate of development for connectivity solutions to keep pace with AI computing will be faster and more essential than ever in the next year.
