AMD’s first fully integrated AI server rack, Helios, has moved into full production and is on track for its initial customer shipments by the end of the third quarter of 2026, the chipmaker confirmed at its Advancing AI 2026 event in San Francisco. Microsoft has already committed to deploying the system “at scale” across Azure, while AMD says volume output will ramp further in the fourth quarter to meet demand stretching into 2027.
AMD announced that its first rack-scale AI system, Helios, is now in full production and will begin shipping to customers by the end of the third quarter of 2026. CEO Lisa Su used the company’s keynote to declare the milestone directly, telling the audience, “Today, Helios has officially entered mass production,” before adding that shipments will begin at the end of the third quarter, with volume supply ramping up in the fourth quarter. Independent analyst firm Moor Insights & Strategy corroborated the milestone, noting AMD put its Helios rack-scale platform into full production with revenue shipments starting at the end of the third quarter.
Inside a Helios Rack: 72 MI455X GPUs and 31TB of HBM4
Helios is built around 72 MI455X GPUs across 18 open-rack, four-GPU compute trays, paired with EPYC 9006 SP7 “Venice” CPUs and AMD Pensando networking. Against Nvidia’s rival Vera Rubin NVL72 system, AMD claims +15% peak FP4, +50% HBM capacity, +6% HBM bandwidth, +50% scale-out bandwidth and up to 30% more tokens per dollar. Pricing has not been officially listed, but industry estimates gathered around AMD’s briefings put a fully configured Helios rack, rated at 2.9 exaflops of FP4 inference performance, priced between $5 million and $5.5 million.
The rack’s memory footprint underlines just how central AI accelerators have become to AMD’s roadmap: HBM4 supply that once might have flowed toward gaming or consumer parts is instead being funnelled into the flagship AI chip. AMD lists an extensive customer roster for the platform, with OpenAI, Meta, Anthropic, Microsoft and Oracle among adopters.
Microsoft Confirms Azure Will “Ramp Helios at Scale”
Days before the Advancing AI keynote, AMD and Microsoft formalised the cloud giant’s commitment. According to AMD’s official announcement, Microsoft will ramp AMD Helios at scale on Azure to power frontier model inference for Microsoft, its AI customers and Azure AI services. The expanded partnership goes beyond GPUs: Azure will add two new VMs powered by 6th Gen AMD EPYC “Venice” processors, alongside deeper integration of AMD’s Pensando networking silicon into Microsoft’s cloud fabric.
Independent coverage of the deal framed it as a watershed for AMD’s data center ambitions. StorageReview noted that the deployment makes Microsoft the first hyperscaler to publicly commit to Helios at scale, the clearest signal yet that AMD’s rack-scale platform is landing as a production alternative to NVIDIA’s rack-scale systems rather than a reference design.
Q4 Ramp Feeds Demand Stretching Into 2027
AMD executives have been careful to frame the Q3 shipment date as only the opening phase. Su’s own comments made clear that volume supply won’t peak until the following quarter, and outside analysts tracking AMD’s supply chain expect the ramp to extend well beyond 2026. That timeline matters because AMD’s data center business is already the company’s growth engine: 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, even before Helios shipments were fully counted.
Read also: AMD Warns of Softer PC Market in H2 2026 Despite Record $11.5B Quarter
What the AI Rack Boom Means for Radeon GPU Buyers
For PC gamers, the Helios story is less about the rack itself and more about where AMD’s chip capacity and cutting-edge memory are being routed. In the same quarter that Data Center revenue exploded, AMD’s Client and Gaming segment grew far more modestly, with Gaming net revenue of $720 million for the three months ended March 28, 2026 increased by 11% compared to net revenue of $647 million for the prior year period, primarily driven by higher demand of our Radeon GPUs. That gap illustrates the widening priority split inside AMD between its AI infrastructure business and its Radeon gaming line.
With TSMC’s advanced packaging capacity and HBM4 allocations increasingly earmarked for MI455X chips destined for Helios racks, gamers in New Zealand and Australia watching for the next generation of Radeon cards may see AMD’s AI commitments shape supply and pricing well before any new gaming GPU reaches local shelves. As hyperscalers like Microsoft lock in multi-gigawatt Helios deployments through 2027, the balance of AMD’s silicon and memory budget looks set to keep tilting toward the data center for the foreseeable future.
Sources
The GamesReviews News Desk covers the latest gaming news across PlayStation, Xbox, Nintendo, PC and esports. Reporting is AI-assisted and reviewed by our editorial team.