Intel plans to ship a new #Artificial Intelligence chip by the end of this year, betting that a cheaper, simpler processor can give it a new foothold in a market dominated by Nvidia, according to a Financial Times interview with Kevork Kechichian, who leads Intel’s data centre group.
The chip, code-named Crescent Island, is built for inference, the work of running AI models, according to the newspaper. Inference is the fastest-growing part of the AI computing market and the area where customers care most about cost. Crescent Island uses lower-cost memory than Nvidia’s chips, and is designed for air-cooled servers, the kind of servers most data centres already run. That makes it easier and cheaper to deploy than chips that need liquid cooling and expensive high-bandwidth memory.
Intel wants to build the chip in its own factories, rather than outsourcing production to Taiwan Semiconductor Manufacturing Company as it has done for many recent products, according to Kechichian. Using its own factories would give Intel more control over supply and could push costs lower than rivals that depend on TSMC for their most advanced chips.
Intel is also assessing whether a version of Crescent Island can be sold in China in compliance with US export controls, according to Kechichian. Washington’s restrictions have squeezed Nvidia out of much of the #🇨🇳 ChinA.I. 🤖🧠🦾🤖 market, and a compliant, lower-cost Intel chip could land in a market its rivals can no longer freely serve.
ft.com/content/3ca15070-c..
