AMD Helios rack system wins Microsoft as customer in Nvidia challenge

Advanced Micro Devices announced Monday that Microsoft will deploy its new Helios rack-scale AI system in Azure data centers, marking a significant expansion of the chipmaker's challenge to Nvidia's infrastructure dominance.

The deal adds Microsoft to a growing roster of Helios customers that includes Meta, OpenAI, and Oracle. AMD plans to begin shipping the systems later this year, according to details first reported by CNBC.

An integrated approach to AI infrastructure

Helios represents AMD's first complete rack-scale system for artificial intelligence workloads, integrating four core technologies the company develops in-house: GPUs, CPUs, networking components, and software. Each of the system's 18 compute trays houses four Instinct GPUs powered by a single EPYC central processing unit, plus up to 12 networking chips based on technology AMD acquired through its 2022 Pensando purchase.

The system directly competes with Nvidia's Grace Blackwell and Vera Rubin offerings. AMD data center head Forrest Norrod told CNBC the company focused on "providing the best total cost of ownership, the lowest cost per token, all in."

The Futurum Group estimates Helios will cost between $5 million and $5.5 million per rack, compared to $3.5 million to $4 million for Nvidia's Vera Rubin system. At up to 7,000 pounds, Helios is also wider and heavier than Nvidia's comparable infrastructure.

Microsoft deployment details

Microsoft CEO Satya Nadella said in a press release that the company is "expanding the Azure infrastructure portfolio with AMD Helios to give customers the performance, scale and choice they need to build and run the next generation of AI applications."

The Helios systems will power frontier model inference for Microsoft and its AI customers while supporting Azure AI services. Microsoft will also introduce two new computing instances running on AMD's latest Venice CPUs—one targeting agentic AI and data pipelines, another for semiconductor design workloads.

The partnership extends a long-standing relationship between the companies, with AMD chips already powering Microsoft's Surface PCs and Xbox consoles. Microsoft was also the first to adopt AMD's MI300X GPU in 2023.

Market position and revenue outlook

Nvidia currently controls more than 95% of the data center GPU market, with AMD holding approximately 4.5%, according to Futurum Group data. However, AMD's data center revenue grew 57% year-over-year in the first quarter of 2026 and now represents the majority of the company's total revenue.

AMD told CNBC it expects to book tens of billions in data center AI revenue starting in 2027, with the majority coming from Helios deployments.

Meta announced in February it would deploy up to 6 gigawatts of AMD GPUs over time, starting with 1 gigawatt on Helios racks later this year. AMD says eight of the top 10 AI companies now run workloads on its Instinct GPUs, including OpenAI, Cohere, and Elon Musk's SpaceXAI.

Why it matters

The addition of Microsoft as a Helios customer validates AMD's integrated infrastructure approach and signals growing enterprise willingness to diversify AI compute suppliers beyond Nvidia. For hyperscalers managing massive AI infrastructure investments, a credible second source reduces supply chain risk and potentially improves pricing leverage. AMD's challenge hinges less on raw chip performance—which analysts say matches Nvidia—and more on whether its ROCm software ecosystem can compete with Nvidia's entrenched CUDA platform in production deployments.

CNBC's Katie Tarasov first reported these details, including exclusive access to AMD's Texas development facility where Helios systems are being tested.