AMD just pulled back the curtain on its next big AI move: a server called Helios, launching in 2026 and powered by its upcoming MI400 series chips.
CEO Lisa Su made the announcement during the Advancing AI event in San Jose, making it clear—this is a shot across the bow at Nvidia’s dominance in AI hardware.
One of the biggest surprises? Sam Altman from OpenAI joined her on stage, confirming that OpenAI is already working with AMD on its future chips, including the MI450 line.
Helios will pack 72 MI400 chips—the same number used in Nvidia’s high-powered NVL72 server. But here’s the twist: AMD isn’t locking down its hardware with proprietary tech like Nvidia’s NVLink. Instead, they’re going open-standard with Helios, aiming to make collaboration easier and more inclusive across the industry.
AMD leaders hammered this point home, saying an open ecosystem is key to AI’s future. That message landed well—Meta, xAI, and Oracle all voiced their support at the event.
Of course, not everything’s rosy. AMD’s ROCm software still trails behind Nvidia’s CUDA when it comes to adoption, but the company’s going all in to catch up. They recently bought server maker ZT Systems and have made 25 AI-related investments in just the past year.
Even with export restrictions to China, AMD is optimistic about AI chip growth. Still, Wall Street wasn’t totally sold—shares dipped 2.2% after the announcement. Analysts pointed out that while the tech is impressive, it probably won’t shake up the market overnight.
