Microsoft Maia 300 Packs 288 GB HBM4, 7x Maia 200 Throughput
TL;DR
- Microsoft plans to deploy over one million Maia 300 chips in 2027, keeping them off the commercial market unlike AWS and Google.
- Each Maia 300 rack delivers 7x the token throughput of Maia 200 and 40% better tokens-per-megawatt versus leading GPU systems.
- Microsoft has secured HBM4 supply from SK Hynix five years out, converting memory procurement from a hardware cycle into a long-term strategic contract.
Microsoft's third-generation in-house accelerator, Maia 300, pairs TSMC N2P compute chiplets with eight HBM4 stacks for 288 GB of capacity and 18 TB/s of bandwidth per chip. Ninety-six of them sit inside a 167 kW liquid-cooled rack, double the accelerator density of the previous Maia 200.
In a deep-dive post on the Azure blog, Microsoft's Rani Borkar and Saurabh Dighe write that the setup delivers "up to 7 times greater rack-scale token throughput in its full 96-accelerator configuration and more than 2.5 times the performance per watt" versus Maia 200. The stated design goal is to "optimize the entire system for LLM inference so that every watt and every dollar delivers more intelligence."
The silicon is a multi-chiplet design. Compute dies ride N2P, while memory-controller and I/O logic sit on separate dies chosen for bandwidth and cost. Layered on top is what Microsoft calls a Software Defined Locally Accessed Dataflow Architecture, giving compilers explicit control over data placement, DMA engines, and synchronization across local memories, compute engines and the network.
At rack scale, 24 four-accelerator trays interconnect electrically inside the rack; an optical second tier can stitch up to 12,288 accelerators into a single superpod across 128 racks, with up to 750 GB/s of cross-rack bandwidth per chip. Liquid cooling absorbs more than 80% of the thermal load. Microsoft names Copilot, GitHub Copilot and Microsoft Foundry as the kind of workloads the system is built around.
The post is architecture, not roadmap. It does not name a production volume, a deployment target in gigawatts, or which products get Maia 300 silicon first — gaps that stand out against the 457 AI infrastructure stories we've tracked over the last 90 days, most of which came attached to a build-out number.
What others are reporting
-
Microsoft Blog Read →
First-party post names GitHub Copilot and Dragon Copilot as first production workloads and details the 96-chip rack configuration and software-defined dataflow architecture.
Maia 300 delivers 40% more tokens per megawatt compared to leading GPU-based systems.
-
Reuters Read →
Reuters independently reports the 1M+ chip deployment volume and anchors the story around cost-savings as Microsoft's primary motivation for replacing third-party silicon.
-
CNBC Read →
Clarifies that Maia 300 is inference-only, keeping Nvidia and AMD in place for training, and frames the financial impact on Microsoft's AI cost structure.
-
The Verge Read →
Tom Warren covers rollout specifics and user-facing product implications for Copilot, giving the announcement its broadest consumer-tech readership framing.
-
The Information Read →
Frames the program as showing signs of life after a slow start, injecting internal execution skepticism that no other outlet's coverage carries.
-
Seoul Economic Daily Read →
Reveals multi-year HBM4 supply contracts and documents high-level Microsoft-SK Hynix meetings, the only coverage to detail the memory supply chain behind the chip.
It is very important for us to establish a high level of trust such that we can place orders.
-
The Tech Portal Read →
Reports the 1M+ chip deployment figure and notes Microsoft is keeping Maia chips off the commercial market, a distinct go-to-market choice versus AWS and Google.
It's basically delivering more intelligence for every watt, every dollar, and every byte of memory by optimizing across the entire system.
-
Crypto Briefing Read →
Provides financial context: Maia represents roughly 10% of Microsoft's AI chip spend by early 2027, with Nvidia and AMD still required for training workloads.
Our biggest bet yet on our own silicon.
Originally reported by azure.microsoft.com
Read the original article →Original headline: Microsoft Unveils Maia 300 Inference Chip on TSMC N2P, Plans Multi-Gigawatt Azure Deployment in 2027