Overview
- Nvidia confirmed its Vera Rubin platform is already in production, touting up to a 10x cut in inference token cost and 4x fewer GPUs for some MoE training versus Blackwell, with hyperscalers such as Microsoft, AWS, Google Cloud, Meta and Oracle preparing adoption in the second half of 2026.
- Nvidia also unveiled Alpamayo for autonomous driving, with the first car using the technology expected to appear on U.S. roads in the first quarter.
- AMD introduced Instinct MI455 for data centers and MI440X for local enterprise use, and previewed MI500 for 2027 with a claimed 1,000x performance leap over an earlier version.
- AMD’s consumer push includes Ryzen AI 400 for on‑device workloads and new Ryzen AI Max chips featuring NPUs up to about 60 TOPS and support for as much as 128 GB of unified memory to run large language models locally.
- Intel announced Core Ultra Series 3 for on‑device AI with reservations starting January 6 and market availability January 27, citing its 18A process, lower power use and integrated graphics improvements for tasks from 4K video to gaming.