Nvidia Corp. has another challenger in AI chip arena, with startup Cerebras Systems, launching a cloud-based AI service based on its chip system this week.
On Tuesday, the Sunnyvale-based company unveiled an inference focused cloud-based service. Cerebras already has a chip and system for AI training. Inference is what AI systems do after they have been trained on vast amounts of data: the systems then make conclusions based on that data. It is expected to become a larger market than the market for training systems.
“We are the fastest in the industry, and the cheapest,” Cerebras co-founder and Chief Executive Andrew Feldman told MarketWatch.
He said the Cerebras cloud service, called Cerebras Inference, is 20 times faster than Nvidia GPU-based solutions in hyperscale clouds. Cerebras is offering pay-as-you-go pricing, for developers and companies. Corporate enterprises can access the service via a Cerebras-managed private cloud or on customer premise.
There is also a free tier with application programming interface (API) access and generous usage limits to anyone who logs in.
Its service is powered by the Cerebras CS-3 system and its AI processor — its Wafer Scale Engine 3 (WSE-3), which is says is the largest chip in the world, focused solely on AI.
The company recently filed confidentially to go public. Feldman declined to comment on the IPO filing.
