As major players in the AI chip market, AMD and NVIDIA have long been recognized for their powerful chips that drive large and energy-intensive AI models. Intel, a long-established name in the computing world, is now stepping up to compete in this arena. Yesterday, during the Intel AI Everywhere launch event in New York, Intel revealed its latest offering, the Gaudi3, an artificial intelligence chip tailored for generative AI software like OpenAI’s ChatGPT, which is currently running on Nvidia GPUs in the cloud.
“We’ve been seeing the excitement with generative AI, the star of the show for 2023,” said Intel CEO Pat Gelsinger At a launch event when unveiling the Gaudi3 along with other chips designed for AI applications, before adding: “We think the AI PC will be the star of the show for the upcoming year.”
Intel’s entry into this market is a response to the dominant position Nvidia has established, highlighted by a 230% increase in Nvidia’s stock year to date, compared to Intel’s 68% rise. AMD, too, has been vying for a piece of this market, with companies increasingly seeking alternatives to Nvidia’s dominance. Gaudi3, set to launch next year, aims to challenge Nvidia’s H100 and AMD’s upcoming MI300X, targeting companies building extensive farms of chips for powering AI applications.
The introduction of Gaudi3 follows Intel’s acquisition of Habana Labs in 2019 and continues its expansion into AI-focused hardware. Alongside Gaudi3, Intel also announced the Core Ultra chips for Windows laptops and PCs, and new fifth-generation Xeon server chips, all featuring a specialized AI component, the NPU, designed for faster AI program execution. However, while these chips won’t match the power needed to run a chatbot like ChatGPT independently, they are adept at smaller tasks, like powering Zoom’s background-blurring feature.
This strategic move is part of Intel’s broader effort to catch up with Taiwan Semiconductor Manufacturing Co. in chip manufacturing prowess, with a target set for 2026. The Core Ultra chips, built using a 7-nanometer process, promise enhanced power efficiency and improved gaming capabilities, boosting the performance of programs like Adobe Premier by over 40%.
Intel’s fifth-generation Xeon processors, often used in tandem with Nvidia GPUs in systems for training and deploying generative AI, are also part of this release. While Intel hasn’t disclosed pricing, these processors are a staple in servers used by large organizations and cloud companies, especially for inferencing tasks in AI deployment, which are less power-intensive than training processes.
With the launch of these new products, Intel is signalling a significant shift in the processor market, aligning its product lines with the growing demand for AI-capable hardware, a move already undertaken by rivals AMD and Qualcomm. This development underscores the evolving landscape of AI technology and its increasing influence on the semiconductor industry.
Featured image: Patrick Gelsinger, chief executive officer of Intel Corp., speaks during the Intel AI Everywhere launch event in New York, US, on Thursday, Dec. 14, 2023. Credit: Victor J. Blue, Bloomberg, Getty Images
