Microsoft to Make Maia Its Primary AI Inference System at Multi-Gigawatt Scale From 2027
Microsoft announced it will scale its in-house Maia chip family to multi-gigawatt deployments and make Maia its primary AI inference system for its own products and services starting in 2027. The company unveiled Maia 300, claiming 40% more tokens per megawatt than leading GPU-based systems, while Maia 200 already serves production workloads including Microsoft Copilot. The move could reduce Microsoft's reliance on Nvidia and AMD hardware, with reported TSMC negotiations for more than 300,000 Maia 300 units for 2027.
IllustrationEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
Microsoft Plans Azure Maia as Primary AI Inference System Starting 2027
Microsoft is expanding its own AI chips in a multi-gigawatt effort aimed at AI inference, positioning Azure Maia as the primary inference system for its own products and services starting in 2027. Maia 200 already serves production workloads, including Microsoft Copilot, running OpenAI models, Microsoft's own MAI models, and open-source models. Maia 300 is set to begin production deployments in 2027, and Microsoft expects it to deliver 40% more tokens per megawatt than leading GPU-based systems. Microsoft does not name Nvidia in the announcement, but the planned rollout could substantially reduce its reliance on Nvidia GPUs for running its own AI services.
Read sourceMicrosoft to Make Maia Chips Primary Inference System, Scaling to Multiple Gigawatts
Satya Nadella officially announced that Microsoft will make its in-house Maia chips the primary inference system for its products, scaling to multiple gigawatts. The move reduces reliance on NVIDIA hardware, with Maia 300 also intended to replace AMD chips. Microsoft claims 40% more tokens per megawatt than leading GPU-based systems. A full rack packs 96 accelerators into 167kW with 27.6TB of HBM, delivering up to 7x the token throughput of a Maia 200 rack. The shift begins in 2027 with Maia 300, a third-generation accelerator for long-context reasoning, low-latency chat and multi-step agent tasks. Its predecessor, Maia 200, already serves production traffic including Microsoft Copilot and reportedly produces up to 1.7x more tokens per watt than leading GPU infrastructure. Maia 300 uses a multi-chiplet design, pairs 8 HBM4 stacks delivering 18.0 TB/s peak bandwidth with a 950W TDP, and supports enhanced MXFP4. Microsoft has reportedly been negotiating with TSMC for more than 300,000 Maia 300 units for 2027 delivery.
Read sourceMicrosoft to Scale Maia Silicon to Multi-Gigawatt Deployments in 2027, Unveils Maia 300
Microsoft announced that its Maia silicon family will scale to multi-gigawatt deployments in 2027, alongside a new Maia 300 chip. According to the post, Maia 300 is described as a complete intelligent system designed for emerging AI workloads including long-context reasoning, latency-sensitive conversational experiences, and sustained multi-step agentic tasks. The post also states that Maia 200 is now serving Microsoft production workloads, including Microsoft Copilot, and is supporting the MAI team in developing next-generation models, with GitHub Copilot and Microsoft Foundry to follow by end of year. The announcement frames the effort as a step toward a fuller full-stack approach.
Read sourceShow 2 older updatesHide older updates
Microsoft Launches Maia 300, Plans Maia as Primary AI Inference System From 2027
Microsoft has launched the Maia 300 and will make Maia its primary AI inference system at multi-gigawatt scale starting in 2027, according to a post on X. The post frames the AI buildout as increasingly centered on inference and on how many tokens can be produced per megawatt. Microsoft cloud chief Scott Guthrie is scheduled to join the poster live at 12:45 p.m. PT / 3:45 p.m. ET on the platform and YouTube. No further technical specifications, pricing, deployment locations, or performance figures were provided in the source item.
Read sourceMicrosoft Unveils Maia 300 Chip and Plans Multi-Gigawatt Maia Inference Expansion
Microsoft announced it will expand its Maia chip family to multi-gigawatt-scale deployment starting next year, positioning the chips as the preferred inference system for its own products and services. The company also released Maia 300, which it says delivers 40% higher token output per watt than mainstream GPU systems. The earlier Maia 200 is already deployed in the United States, Europe, and Australia, where Microsoft reports up to 1.7x higher token output per watt in Teams and Copilot scenarios. The announcement, made by CEO Satya Nadella, outlines Microsoft's self-developed chip roadmap and provides energy-efficiency figures for Maia 300 and Maia 200, offering a basis for comparing trade-offs against mainstream GPU solutions.