Intel’s Xeon Revolution: Inferencing Crown Jewels Challenge Nvidia’s AI Throne

At the “AI Everywhere” event held in New York, USA, Intel introduced a suite of Artificial Intelligence (AI) products, including the Intel Core Ultra mobile processor and the fifth-generation Intel Xeon processor. Intel CEO Pat Gelsinger also debuted the Intel Gaudi 3 accelerator, set to be released on schedule next year.

According to Tom’s Hardware, during the event, Pat Gelsinger critiqued NVIDIA’s CUDA ecosystem, asserting that inferencing technology will be more crucial than AI training. Responding to media queries, Gelsinger opined that NVIDIA’s reliance on CUDA for AI training would not sustain its advantage indefinitely.

An illustration shows the assembly process of the Intel Core Ultra processor. On Dec. 14, 2023, Intel introduced the Intel Core Ultra mobile processor family. Powered by Intel’s 3D performance hybrid architecture and built on the Intel 4 process, new H- and U-Series processors deliver a balance of performance and power efficiency, immersive experiences, and AI acceleration. (Credit: Intel Corporation)

Gelsinger highlighted that the entire industry is actively moving away from CUDA, citing examples of MLIR, Google, and OpenAI, suggesting a shift towards making AI training more open and accessible. The industry is motivated to introduce broader technologies for training, innovation, and data science. Intel’s focus isn’t limited to just AI training; instead, inferencing is deemed the right approach.

Additionally, Gelsinger took this opportunity to promote OpenVINO, Intel’s standard for its AI initiatives, which fosters a world of hybrid computing. Intel is committed to being 100% involved in the AI competition in data centers and is exploring every business opportunity, including collaborations with companies like NVIDIA and AMD, aiming to become a key player in the semiconductor foundry space.