Companies/Microsoft
Redmond, WA, USA
Microsoft
Azure operates some of the largest GPU clusters in existence, only a fraction of which have ever submitted HPL results.
Systems this company appears in
- Systems
- 20
- Aggregate Rmax
- 757.6 PFlop/s
- Aggregate power
- 359 kW
- First appearance
- 2002
- Most recent
- 2026-06-22
- Concentration
- 74%
Aggregate Rmax is the sum of the systems this company supplies any part of. It is not revenue and not attributable share: an integrator and a memory vendor in the same machine both count its full FLOPS here. Attribution weighted by role is Phase 4.
Role breakdown
| Role | Systems | Rmax in those systems |
|---|---|---|
| Integrator | 14 | 757.2 PFlop/s |
| Operating system | 6 | 433.91 TFlop/s |
Appearances
| System | Supplied as | Shipped under | Rmax | Rank |
|---|---|---|---|---|
| Azure NDv6 GB300 Cluster USA · 2025-10 | Integrator | Microsoft | - | unranked |
| Met Office Azure Supercomputer United Kingdom · 2025-05 | Integrator | Microsoft | - | unranked |
| Reindeer USA · 2024 | Integrator | Microsoft | 45.59 PFlop/s | #32 November 2024 → |
| Eagle USA · 2023-11 | Integrator | Microsoft | 561.2 PFlop/s | #3 November 2023 → |
| Microsoft Goodyear AI Datacenter USA · 2023-04 | Integrator | Microsoft | - | unranked |
| Explorer-WUS3 USA · 2023 | Integrator | Microsoft | 53.96 PFlop/s | #48 June 2026 → |
| Microsoft Project Osmium USA · 2021-01 | Integrator | Microsoft | - | unranked |
| Voyager-EUS2 USA · 2021 | Integrator | Microsoft | 30.05 PFlop/s | #10 November 2021 → |
| Pioneer-SCUS USA · 2021 | Integrator | Microsoft | 16.59 PFlop/s | #27 June 2021 → |
| Pioneer-WEU Netherlands · 2021 | Integrator | Microsoft | 16.59 PFlop/s | #28 June 2021 → |
| Pioneer-WUS2 USA · 2021 | Integrator | Microsoft | 16.59 PFlop/s | #35 November 2021 → |
| Pioneer-EUS USA · 2020 | Integrator | Microsoft | 16.59 PFlop/s | #26 June 2021 → |
| Faenov (Windows Azure) USA · 2012 | Operating system | Microsoft | 151.30 TFlop/s | #165 November 2012 → |
| Baku (HLRS) Germany · 2009 | Operating system | Microsoft | 50.79 TFlop/s | #77 June 2009 → |
| Akka Sweden · 2008-06-25 | Operating system | Microsoft | 46.04 TFlop/s | #40 June 2008 → |
| Magic Cube China · 2008 | Operating system | Microsoft | 180.60 TFlop/s | #11 November 2008 → |
| Lincoln (NCSA Windows) USA · 2006 | Operating system | Microsoft | 4.11 TFlop/s | #129 June 2006 → |
| Cornell Velocity2 USA · 2002 | Operating system | Microsoft | 1.07 TFlop/s | #49 June 2003 → |
| Microsoft Azure AI supercomputer for OpenAI (2020) USA · | Integrator | Microsoft | - | unranked |
| Microsoft Pecos Campus USA · | Integrator | Microsoft | - | unranked |
Parts supplied
| Part | Category | Generation | Released |
|---|---|---|---|
| Microsoft Maia 200 | Accelerators | Maia | 2026 |
| Microsoft Maia 100 | Accelerators | Maia | 2023 |
| Windows HPC Server 2008 | Operating systems | Windows HPC Server | 2008 |
AI datacenters operated
| System | Status | Accelerators | Stated power (MW) |
|---|---|---|---|
| Eagle | operational | 14,400 | - |
| Microsoft Fairwater (Atlanta) | operational | 304,279 | 859 |
| Microsoft Fairwater (Wisconsin) | operational | 176,391 | 442 |
| Microsoft Azure AI supercomputer for OpenAI (2020) | operational | 10,000 | - |
| Microsoft Azure AI supercomputer (Loughton, UK) | under construction | 23,040 | - |
| Microsoft Azure AI datacenter (Narvik, Norway) | under construction | 30,000 | 230 |
| SINES Data Campus GB300 Deployment | operational | 12,600 | - |
| Microsoft Project Osmium | operational | 103,200 | 190 |
| Microsoft Goodyear AI Datacenter | operational | 116,400 | 202 |
| Hypercluster (Microsoft Research) | operational | - | - |
Figures come from the AI datacenter metrics; blank means none recorded.
Compute deals
| Date | Role | Counterparty | Type | MW | Value ($bn) | Source |
|---|---|---|---|---|---|---|
| 2026-05-05 | Buyer | Nscale | Cloud contract | 200 | - | nscale.com |
| 2026-04-14 | Buyer | Nscale | Cloud contract | 230 | - | nscale.comdatacenter-forum.com |
| 2025-11-18 | Seller | Anthropic | Cloud contract | 1,000 | 30 | anthropic.comblogs.microsoft.com |
| 2025-11-03 | Buyer | G42 | Other | - | 15.2 | blogs.microsoft.com |
| 2025-11-03 | Buyer | IREN | Cloud contract | 200 | 9.7 | sec.govsec.gov |
| 2025-10-28 | Seller | OpenAI | Cloud contract | - | 250 | blogs.microsoft.com |
| 2025-10-15 | Buyer | Nscale | Cloud contract | - | - | nscale.comlp.startcampus.pt |
| 2025-09-17 | Buyer | Nscale | Cloud contract | - | 6.2 | filingreader.comnews.microsoft.com |
| 2025-09-08 | Buyer | Nebius | Cloud contract | - | 17.4 | sec.gov |
Headline values are as announced and are not additive. See all compute deals.
Quantum computers operated
- Microsoft Majorana 1 (Other, operational, 8 qubits)
- Microsoft Majorana 2 (Other, operational)
Sites operated
| Site | Location | Cooling | Commissioned |
|---|---|---|---|
| Microsoft Azure (undisclosed US region) | USA | - | 2023 |
| Microsoft Mount Pleasant AI datacenter site | Mount Pleasant, WI, USA | Closed-loop liquid cooling | 2025 |
| Microsoft Atlanta AI datacenter site | Atlanta, GA, USA | Closed-loop liquid cooling | 2025 |
| Microsoft Cumming Datacenter Campus | Cumming, IA, USA | - | - |
| Microsoft Goodyear Datacenter Campus | Goodyear, AZ, USA | - | - |
| Microsoft Pecos Site | Pecos, TX, USA | Closed-loop, no steady-state water use | - |
Related news
- Microsoft Boosts HBM Bandwidth by 2.5x with Maia 300 Accelerator · HPCwireMicrosoft's own technical brief lays out Maia 300 succeeding Maia 200: a shift from a monolithic 3nm die to a chiplet design with NUMA memory access and a two-tier scale-up network that supports superpods of over 12,000 accelerators, roughly doubling Maia 200's largest cluster size. It's a second generation of in-house Azure inference silicon, relevant to how much of Microsoft's AI fleet runs on custom ASICs rather than merchant GPUs.
- Microsoft files to build data center campus in Prince William County, Virginia · Data Center DynamicsMicrosoft filed federal and state permits for a two-building data center campus on 123 acres it bought for $465.5M in 2024, directly bordering an Amazon campus and near a Google-linked site in the same Gainesville corridor. It's another data point on how concentrated the current AI buildout is in a handful of Northern Virginia counties, with the largest hyperscalers now building adjacent to rather than in competition for the scarce sites there.
- Microsoft targets 38GW of data center capacity in 2032 - report · Data Center DynamicsMicrosoft's reported target of roughly tripling to 38GW of capacity by 2032, from about 12GW now, is a forward plan sourced to Bloomberg's unnamed sources rather than a confirmed figure, so it belongs here as stated intent rather than anything checkable against installed hardware. The structural point for the graph is how much of that growth is meant to come from leasing neocloud capacity (CoreWeave, Nscale, Lambda, Iren, Nebius were named) rather than built and owned outright, which is exactly the operator/owner distinction entity resolution here has to keep straight rather than collapse into one Microsoft footprint.
- Microsoft links Fairwater Atlanta and Wisconsin into one "AI superfactory" · MicrosoftThe two Fairwater sites are joined over a dedicated AI WAN fiber backbone so training jobs can span both facilities as one system rather than two separate clusters, a different scaling strategy than the single site gigawatt campuses xAI and Meta are building. Each rack carries up to 72 NVIDIA Blackwell GPUs; Microsoft has not disclosed a total GPU count for either site.
Further reading
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.
Programming a supercomputer
The software stack beneath the application, and why the OS is now a sovereignty signal as much as a technical choice.
Inside a node
Why accelerators exist, what the memory hierarchy costs, and how the CPU and accelerator merged onto one package.