Nvidia and AMD Intensify AI Chip Race with Next-Gen Platforms

The competition between Nvidia and AMD in the artificial intelligence hardware market is heating up, with both companies vying for dominance in the rapidly expanding AI data center space. Recent developments suggest Nvidia may gain a temporary lead with its Vera Rubin platform, while AMD is working to maintain its timeline for the Helios system, despite reports of potential delays.

Nvidia’s Vera Rubin Platform Gains Momentum

Nvidia’s next-generation Vera Rubin platform is potentially on track for an earlier-than-anticipated release, possibly as early as the second quarter of 2026, according to reports from Evercore ISI. This is driven by the fact that silicon for the platform is already in mass production. Jensen Huang stated in early January that the Vera Rubin platform was already in production, potentially allowing key customers to receive the new AI platform ahead of schedule.

To maintain its competitive edge, Nvidia has upgraded the specifications of its Vera Rubin NVL72 AI server, increasing HBM4 memory bandwidth by 10% to 22.2TB/sec, surpassing AMD’s Instinct MI455X as reported by Tweaktown. This enhancement utilizes faster 8-Hi HBM4 stacks.

AMD Addresses Delay Reports for Helios System

Reports from SemiAnalysis initially suggested potential delays in the production and adoption of AMD’s next-generation Instinct MI455X. However, AMD swiftly refuted these claims, stating that the platform remains “on target for 2H 2026”, as confirmed by Anush Elangovan, corporate vice president of AMD’s software development, via a post on X.

AMD’s Helios rack-scale solutions for AI are designed to pack 72 Instinct MI455X AI accelerators with 31 TB of HBM4 memory, aiming to deliver 2.9 FP4 exaFLOPS for AI inference, and 1.4 FP8 exaFLOPS for AI training. Initial plans for UALink interconnections for scale-up connectivity have reportedly shifted to UALink over Ethernet for initial Helios machines, potentially impacting performance.

Astera Labs, a key developer of connectivity solutions, anticipates ramps for UALink-based platforms in 2027, not 2026 according to Tom’s Hardware.

Competitive Dynamics and Market Implications

If Nvidia successfully accelerates the release of its NVL72 VR200 platform while AMD experiences delays in ramping up its Helios solution, Nvidia could strengthen its leadership in the AI market. Developers of advanced AI models are currently heavily reliant on Nvidia’s hardware as noted by Wccftech. The competitive pressure is driving both companies to innovate and push the boundaries of AI hardware capabilities.

Nvidia’s recent upgrades to the Vera Rubin platform are seen as a direct response to AMD’s pursuit of the Instinct MI455X AI chips according to reports from Wccftech.

Key Takeaways

• Nvidia’s Vera Rubin platform is potentially launching ahead of schedule, possibly in Q2 2026.
• Nvidia has increased the HBM4 memory bandwidth of its Vera Rubin NVL72 server to 22.2TB/sec, surpassing AMD’s MI455X.
• AMD refutes reports of significant delays to its Helios system, maintaining a 2H 2026 target.
• Competition between Nvidia and AMD is driving rapid innovation in AI hardware.