Reports say AMD's first rack-scale AI system "Helios" mass production has been delayed until 2027

2026-02-19 02:39:20

Tech Home, February 17 — Semiconductor analysis firm SemiAnalysis stated in its report today that AMD’s first rack-scale AI system, MI455X UALoE72 “Helios,” has encountered manufacturing delays.

According to the report, the MI455X UALoE72 will begin engineering sample production and small-scale mass production in the second half of 2026, while the first batch of token generation for large-scale mass production and commercial deployment will not occur until the second half of 2027. This means that “Helios” will, to some extent, compete with NVIDIA’s “Rubin” and even the “Rubin Ultra” platform.

AMD will implement Ethernet-based UALink high-speed interconnect in “Helios” to create an integrated rack-scale deployment solution with a high number of XPU units, aiming to catch up with NVIDIA (GPU), Google (TPU), and Amazon AWS (Trainium), who are already ahead in this area.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

1 Likes