AMD MI455 Debut at CES Signals AI Push
AMD MI455 unveiled at CES with Ryzen AI 400 and Helios rack; OpenAI-linked deals boost near-term sales expectations and data-center positioning.
KEY TAKEAWAYS
- OpenAI agreement signed October 2025 ties MI400 rollout to billions in annual revenue.
- MI455 boosts capacity with 70.0% more transistors and 400GB HBM4 memory.
- Helios rack delivers up to 2.9 exaflops per rack for customers beyond hyperscalers.
HIGH POTENTIAL TRADES SENT DIRECTLY TO YOUR INBOX
Add your email to receive our free daily newsletter. No spam, unsubscribe anytime.
Advanced Micro Devices Inc. (AMD) unveiled the MI455 GPU at CES on Jan. 5, 2026, alongside the Ryzen AI 400 processors and a Helios rack delivering up to 2.9 exaflops. The announcements position AMD to expand AI compute across AI PCs and smaller corporate data centers while previewing an MI500 roadmap for 2027.
Chip Specifications and Roadmap
The MI455 is AMD’s next MI-series data-center accelerator, featuring 70% more transistors than the MI355 and 400GB of HBM4 high-bandwidth memory. These upgrades aim to boost model throughput and support larger working sets at the accelerator level.
AMD also introduced the Ryzen AI 400 family for consumer and prosumer machines, highlighting improvements in multitasking, creative workflows, overall performance, and memory compared with prior chips. The lineup includes the Ryzen AI Max+ variant, designed for local inference and gaming.
Executives previewed the MI500, expected to launch in 2027, which the company said will deliver roughly 1,000 times the performance of an older processor version. This benchmark signals AMD’s multi-year ambition in AI compute.
The MI440X, an enterprise version of the MI400 family, targets on-premise, non–AI-specific infrastructure. AMD tied the MI400 series to initial OpenAI deployments scheduled for 2026, which the company expects will drive significant sales growth.
Partnerships and Commercial Outlook
AMD framed the product rollouts around commercial deals and enterprise demand. Its October 2025 agreement with OpenAI, described as adding billions in annual revenue, was highlighted during the CES presentation. OpenAI President Greg Brockman joined the event to emphasize the scale of computing needs.
The company also expanded its partnership with Luma AI for video-generation workloads. Luma’s chief executive praised AMD’s ROCm software and the platform’s total-cost-of-ownership profile. Infrastructure products like the Helios rack, capable of up to 2.9 exaflops per rack, aim to serve customers beyond hyperscalers.
Together, the chipset advances, OpenAI-linked commercial commitments, and the MI500 roadmap connect AMD’s announcements to near-term sales opportunities and a longer-term effort to compete for AI workloads outside the largest cloud providers.
