IBM is considering using its in-house designed artificial intelligence chips to reduce the cost of cloud computing services.
The company’s general manager, Kumar Kare, told Reuters in an interview at the San Francisco Semiconductor Conference that the company is considering a chip called an “artificial intelligence unit” as part of its new “WatsonX” cloud. part of the service.
IBM hopes to provide users with more convenient services using generative artificial intelligence technology, which can better learn human text and language schemes than Watson, its first major artificial intelligence system launched more than a decade ago. .
It is said that one of the main obstacles currently facing the old “Watson” system is the high cost, and IBM hopes to use its own AI chip to solve this problem this time. Carre said that using its own chips can reduce the cost of cloud computing services because they are very energy efficient.
IBM unveiled its AI chip in October last year, but did not disclose details such as the manufacturer or how it will be used. The chip is manufactured by Samsung Electronics, which has a semiconductor research partnership with IBM and is being considered by IBM for Watson X, Carre said.
IBM has yet to say when its chips will be available to cloud customers, but Carre said the company already has thousands of prototypes in operation.
Other tech giants, such as Alphabet’s Google and Amazon, have already begun designing their own AI chips, it was previously reported.
But Carre said IBM wasn’t trying to design a direct replacement for Nvidia’s chips, which have a lead in training large-scale AI systems with vast amounts of data. Instead, IBM’s chip aims to be cost-effective in what industry insiders call inference, the process of using already trained AI systems for actual decision-making.
