Intel bets on right-sized AI systems, not the biggest ones, at OCP

At the Open Compute Project Global Summit, Intel detailed its GPU D-series (formerly Crescent Island): 1.3 petaflops of FP8, 1.5 TB/s of memory bandwidth, 350W air-cooled and up to 480GB of LPDDR5X. The company argues inference and agentic AI will bring CPUs back to the center of the data center.

• D-series: 1.3 PFLOPS FP8, 1.5 TB/s, 350W, up to 480GB LPDDR5X
• Supports up to 300B-parameter models on one card, 1T+ on eight GPUs
• Xeon ORv3 racks scale beyond 46,000 cores at 133.45 kW
• Open-weight models' share of token traffic rose from 31% to 78%