“The CPU is reinserting itself”, will reach parity with GPUs for AI workloads: Tan

AI workloads are requiring more and more CPUs to handle the likes of communications, memory paging, interconnect orchestration, and distributed scheduling, says Intel. CEO Lip-Bu Tan insists that this view is “not just our wishful thinking” and that CPU and GPU use will soon reach parity.

When it comes to inference, “the ratio of CPU to GPU used to be 1:8, and now it's 1:4 and I think [moving] towards parity or even better,” Tan told analysts on a Q1 call. The chipmaker reported a $3.7 billion loss – but showed some signs of green shoots, which sent stocks up 24% on Friday.

After a grim few years of heavy restructuring, executives were upbeat after beating revenue guidance. Intel is now “data-driven, paranoid and engineering-centric” said Tan, pointing to “seen clear signs that the CPU is reinserting itself as the indispensable foundation of the AI era…”

Markets seemed to agree, with shares soaring, despite the huge increase in losses: investors liked improved Intel Foundry performance, demand outstripping supply (“demand continues to run ahead of supply for all our businesses, especially for Xeon server CPUs”) and expanding chip TAM.