Cerebras Systems launches 3rd-generation wafer-scale AI processor and new AI server, claiming fastest AI chip globally.

Cerebras Systems has introduced the WSE-3, a new 5-nanometer AI processor that the Sunnyvale, California-based startup claims to be the fastest in the world. The WSE-3 chip doubles the performance of its predecessor, according to the company announcement made on March 13. "Once again, we've delivered the biggest and fastest AI chip on the planet with the same dinner plate-size form factor," said Andy Hock, Cerebras' vice president of product management. The new processor powers Cerebras' CS-3 AI server, which is designed to train the largest AI models.

Cerebras entered the hardware market in 2019 with the original Wafer Scale Engine (WSE), an eight-by-eight-inch chip that was 56 times larger than the largest GPU at the time, featuring 1.2 trillion transistors and 400,000 computing cores. In 2021, the company launched the WSE-2, a 7-nanometer chip that doubled performance with 2.6 trillion transistors and 850,000 cores. The WSE-3 features four million transistors and 900,000 cores, delivering 125 petaflops of performance. "The CS-3 is a big step forward for us," Hock told Data Center Knowledge. "It's two times more performance than our CS-2 server. So, it's two times faster training for large AI models with the same power draw, and it's available at the same price to our customers."

Cerebras has positioned itself as an alternative to Nvidia GPU-powered AI systems, with the pitch that customers can run AI training on Cerebras hardware using significantly fewer chips than traditional approaches. "One Cerebras server can do the same work as 10 racks of GPUs," said Karl Freund, founder and principal analyst of Cambrian AI Research. Nvidia dominates the AI market with its GPUs capturing about 85% of the AI chip market, while remaining players such as AMD, Intel, Google, AWS, Microsoft, Cerebras, and others have captured about 15%. The company generates revenue by selling servers that cost millions of dollars each, making its CS-3 systems available to customers over the cloud, to large enterprises, government agencies, and international cloud providers. Recent customers include Mayo Clinic, Argonne National Laboratory, and pharmaceutical giant GlaxoSmithKline.

In July 2023, Cerebras announced a $100 million deal with G42, a technology holding group based in the United Arab Emirates, to build the first of nine interconnected, cloud-based AI supercomputers. Since then, the two companies have built two supercomputers totaling eight exaflops of AI compute, accessible over the cloud and optimized for training large language models and generative AI models. Cerebras and G42 are currently building the Condor Galaxy 3 supercomputer in Dallas, which will be powered by 64 CS-3 systems and will produce eight exaflops of AI compute. By the end of 2024, the companies plan to complete all nine supercomputers, which will total 55.6 exaflops of compute.

While Cerebras has found success as an early player in the AI chip sector, the company addresses the very high end of the market and is not a high-volume product. "From the beginning, Cerebras took a very different approach," Freund said. "Everybody else is trying to outdo Nvidia, which is really hard to do. Cerebras said, 'We're going to build an entire wafer-scale AI engine,' which no one has ever done. The benefit is incredibly high performance." According to analyst Matt Kimball at Moor Insights & Strategy, "It's not like an HPE or Dell product where everybody – from a small business to large enterprise – buys it. Smaller companies are slower to adopt new technology, and I suspect this may be cost prohibitive."