IBM mulls using its own AI chip in new cloud service to lower costs
IBM is thinking about using its own artificial intelligence chips to reduce the expenses of running a cloud computing service that it just released, a company representative mentioned on Tuesday.
During a discussion with Reuters at a semiconductors meetup held in San Francisco, Mukesh Khare, the head honcho of IBM Semiconductors, disclosed that they are currently pondering over the integration of a chip known as the Artificial Intelligence Unit within their latest cloud service termed "watsonx".
IBM is aiming to capitalize on the surge in generative AI technologies capable of producing human-like written content, following the lackluster performance of Watson, its initial significant AI system, over ten years ago.
IBM hopes to overcome a challenge faced by the previous Watson system with regards to its exorbitant expenses. Khare mentions that by utilizing their own chips, the company aims to reduce the costs associated with cloud services, as these chips possess exceptional energy efficiency.
IBM made public the chip's presence in October but did not reveal the producer or its intended application.
According to Khare, a representative from Samsung Electronics, they have produced the chip in collaboration with IBM for their semiconductor research. He also mentioned that his company is contemplating utilizing it for watsonx.
IBM does not have a fixed timetable for when the chip might be accessible for utilization by cloud clients, but according to Khare, the organization has numerous thousand initial models of chips that are already functioning.
IBM has partnered with other prominent tech companies, including Google from Alphabet and Amazon.com, to develop its own AI processors.
However, Khare clarified that IBM is not attempting to create a direct substitute for Nvidia's semiconductors, which currently dominate the market in powering AI systems that necessitate immense volumes of data for training.
On the contrary, IBM's chip seeks to be economical in the realm of artificial intelligence, specifically in inference. Inference refers to utilizing a pre-trained AI system to make practical decisions in the real world, as per experts in the AI industry.
"That is currently where the main focus lies," Khare commented. "At the moment, we do not intend to shift our attention towards training. Training involves a different level of computational resources. Our aim is to invest in areas where we can make the largest positive influence." (Reported by Stephen Nellis; Edited by Jamie Freed)
