AMD has moved its Helios rack-scale AI system into full production, with the company confirming that first customer shipments are on track for the end of Q3 2026. Chair and CEO Dr. Lisa Su announced the milestone during the Advancing AI 2026 keynote in San Francisco, where Microsoft was unveiled as the newest hyperscaler committing to deploy the platform on Azure.

AMD’s CEO, Lisa Su, announced that production of Helios has started, with shipments scheduled for the end of Q3 2026. The disclosure caps roughly a year of previews and partner announcements for the system, which AMD first teased in 2025 before showing it publicly at CES 2026 in January.

Inside a Helios Rack: 72 GPUs, EPYC CPUs and 31TB of HBM4

According to AMD’s own event materials, Helios pairs 72 Instinct MI455X GPUs with 6th Gen EPYC CPUs and Pensando networking over open UALink-over-Ethernet, rated at 2.9 exaflops of FP4 compute. AMD’s Santa Clara-based data center press office separately confirmed the platform is built to combine AMD Instinct MI455X GPUs, AMD EPYC “Venice” CPUs, Pensando networking and ROCm software in an open, integrated rackscale platform built for large-scale AI training and inference.

Analysts tracking the keynote pegged pricing for a fully configured rack at roughly $5 million to $5.5 million, positioning Helios directly against Nvidia’s Vera Rubin and Grace Blackwell rack systems rather than against individual GPU cards. TechCrunch reported that Su called Helios the tech industry’s “highest-performance AI rack,” adding that it was “built to train and run the most demanding frontier models in the world at massive scale.”

Microsoft Joins Meta, OpenAI, Oracle and Anthropic as Customers

In its own newsroom statement, AMD confirmed that at the center of this expansion, Microsoft will deploy the AMD Helios Rackscale Solution, to power frontier model AI inference for Microsoft, its AI customers and support Azure AI services. Su framed the tie-up as a natural extension of an existing relationship, saying “AMD and Microsoft have spent years building high-performance infrastructure together, and today we’re extending that partnership across the full stack of AMD AI solutions on Azure.”

Microsoft is far from the only name on AMD’s customer sheet. TechCrunch noted that Helios, which was revealed in 2025 and shown onstage in January at CES 2026, already has several well-known customers, including OpenAI, Meta, Oracle, Anthropic, and Microsoft, all of which have plans to deploy the system. Separately, AMD and Anthropic announced a strategic partnership under which Anthropic and AMD announced a strategic partnership to deploy up to two gigawatts of GPUs via the new rack system.

CNBC framed Microsoft’s arrival as evidence AMD finally has a credible rival product to Nvidia’s dominant rack platforms, noting it’s the first rival to Nvidia’s wildly popular Grace Blackwell and Vera Rubin systems, and is aiming to give the world’s most valuable chipmaker its first real competition in years.

Shipments Ramp From Q3 Into Q4, With 2027 Framed as the Real Payoff

Industry analyst firm Moor Insights & Strategy, which attended the keynote, singled out the timeline as the standout line of the event: the single most important sentence of the keynote was Su declaring Helios in full production, with shipments starting at the end of Q3 and ramping through Q4. The firm also noted AMD’s broader roadmap ambitions, including MI500 in 2027 with copper and optical interconnect options, MI600 in 2028, a new Helios every year, and a claimed 2,000x-plus inference gain over four years.

AMD is explicit that the real financial payoff arrives after this year’s ramp. The company anticipates generating tens of billions of dollars in data center AI revenue from 2027 onward, with Helios expected to be the primary driver. That target sits on top of already fast-growing data center numbers: AMD’s most recent quarterly filing shows Data Center segment revenue was $6.7 billion, up 107% year-over-year, driven by strong demand for AMD EPYC processors and AMD Instinct GPUs.

Why a Datacenter Rack Story Matters to PC and Console Gamers

Helios itself never touches a gaming rig — it’s built for cloud-scale AI training and inference, not Radeon graphics cards. But the same Q2 filing that detailed the Data Center surge also showed AMD’s gaming-adjacent business holding its own, with the company reporting expanded global availability of AMD Radeon RX 9070 GRE graphics cards, extending performance and advanced ray tracing capabilities to more gamers, while Client and Gaming segment revenue was $3.8 billion, up 6% year-over-year. That matters because AMD has repeatedly signalled data center is now its priority business, which raises the perennial question of whether the AI boom crowds out fab capacity and R&D dollars that would otherwise go to Radeon and console silicon.

There’s also a genuine regional angle for Australian and New Zealand readers. Microsoft’s own Azure global infrastructure listing confirms Azure is available or coming soon to Australia East, Australia Southeast, Australia Central, Australia Central 2, and New Zealand North regions, meaning the Helios racks Microsoft is now committing to could eventually feed inference capacity into services many local gamers, streamers and developers already touch through Azure-hosted tools, Xbox Cloud backend infrastructure and AI-assisted game services. As AMD ramps Helios through Q4 and chases OpenAI, Meta and Anthropic’s multi-gigawatt commitments into 2027, the knock-on effects for silicon supply, cloud pricing and even future Radeon roadmaps are worth watching from this side of the Pacific.

Read also: AMD Braces for Softer PC Market in H2 2026, Still Sees Ryzen Beating the Slump

Sources