International Business Machines (IBM) is considering using artificial intelligence chips designed in-house to lower the costs of operating a cloud computing service it made widely available this week, an executive said Tuesday.
The company is contemplating using a chip called the Artificial Intelligence Unit as part of its new “watsonx” cloud service, according to Mukesh Khare, general manager of IBM Semiconductors, in an interview with Reuters at a semiconductor conference in San Francisco.
More than a decade after Watson, its first major AI system struggled to find a market, IBM hopes to benefit from the boom in generative AI technologies that can write human-like text.
High costs were one of the problems the previous Watson system had, which IBM aims to fix now. Because its chips are exceedingly power-efficient, Khare claimed doing so could reduce the cost of cloud services.
In October, IBM made the chip’s existence announcement but did not specify the manufacturer or how it would be used.
According to Khare, his company is considering using the chip in watsonx. It is manufactured by Samsung Electronics, which collaborates with IBM on semiconductor research.
IBM has joined other tech giants such as Alphabet’s Google and Amazon.com in designing its own AI chips.
However, Khare claimed that IBM was not trying to design a direct replacement for Nvidia semiconductors, whose chips lead the market for using massive amounts of data to train AI systems.
Instead, IBM’s chip aims to be cost-effective at a process known as inference, which uses a trained AI system to make decisions in the real world.
“That’s where the volume is right now,” Khare said. “We don’t want to go toward training right now. Training is a different beast in terms of computing. We want to go where we can have the most impact.”