Intel CEO, Pat Gelsinger, aggressively criticized Nvidia’s CUDA technology, asserting that inference technology will surpass training for AI. He made these remarks while unveiling the Intel Core Ultra and 5th Gen Xeon data center chips at an event in New York City. During a Q&A session at the NASDAQ, Gelsinger implied that Nvidia’s current dominance in AI training through CUDA is not permanent.
In a statement, Intel CEO expressed a strong determination to challenge the CUDA market dominance. Gelsinger pointed to initiatives by MLIR, Google, and OpenAI as evidence of a shift towards a more open approach to AI training, emphasizing the industry’s push for innovation and broader technology adoption beyond CUDA.
Despite the focus on training, Intel is prioritizing inference. The Intel CEO indicated that Intel is prepared to meet this challenge with the introduction of Gaudi 3, which was showcased for the first time. Additionally, Intel aims to achieve this capability with Xeon and edge PCs.
The Intel CEO also used the occasion to promote OpenVINO, the standard that Intel has rallied around for its AI initiatives. He also predicted a future of mixed computing, with some processes taking place in the cloud and others on personal computers.
Sandra Rivera, the executive vice president and general manager of Intel’s Data Center and AI Group, highlighted that Intel’s extensive reach from the data center to the PC could position it as a preferred partner due to its capacity for large-scale production.