This week we’ve got launched a wave of purpose-built datacenters and infrastructure investments we’re making around the globe to assist the worldwide adoption of cutting-edge AI workloads and cloud companies.
Right now in Wisconsin we launched Fairwater, our latest US AI datacenter, the most important and most subtle AI manufacturing facility we’ve constructed but. Along with our Fairwater datacenter in Wisconsin, we even have a number of equivalent Fairwater datacenters beneath development in different areas throughout the US.
In Narvik, Norway, Microsoft introduced plans with nScale and Aker JV to develop a brand new hyperscale AI datacenter.
In Loughton, UK, we introduced a partnership with nScale to construct the UK’s largest supercomputer to assist companies within the UK.
These AI datacenters are vital capital tasks, representing tens of billions of {dollars} of investments and tons of of 1000’s of cutting-edge AI chips, and can seamlessly join with our world Microsoft Cloud of over 400 datacenters in 70 areas around the globe. By means of innovation that may allow us to hyperlink these AI datacenters in a distributed community, we multiply the effectivity and compute in an exponential approach to additional democratize entry to AI companies globally.
So what’s an AI datacenter?
The AI datacenter: the brand new manufacturing facility of the AI period
An AI datacenter is a singular, purpose-built facility designed particularly for AI coaching in addition to working large-scale synthetic intelligence fashions and functions. Microsoft’s AI datacenters energy OpenAI, Microsoft AI, our Copilot capabilities and lots of extra main AI workloads.
The brand new Fairwater AI datacenter in Wisconsin stands as a outstanding feat of engineering, masking 315 acres and housing three large buildings with a mixed 1.2 million sq. toes beneath roofs. Establishing this facility required 46.6 miles of deep basis piles, 26.5 million kilos of structural metal, 120 miles of medium-voltage underground cable and 72.6 miles of mechanical piping.
In contrast to typical cloud datacenters, that are optimized to run many smaller, unbiased workloads corresponding to internet hosting web sites, e-mail or enterprise functions, this datacenter is constructed to work as one large AI supercomputer utilizing a single flat networking interconnecting tons of of 1000’s of the newest NVIDIA GPUs. The truth is, it’s going to ship 10X the efficiency of the world’s quickest supercomputer at this time, enabling AI coaching and inference workloads at a stage by no means earlier than seen.
The position of our AI datacenters – powering frontier AI
Efficient AI fashions depend on 1000’s of computer systems working collectively, powered by GPUs, or specialised AI accelerators, to course of large concurrent mathematical computations. They’re interconnected with extraordinarily quick networks to allow them to share outcomes immediately, and all of that is supported by monumental storage techniques that maintain the info (like textual content, photographs or video) damaged down into tokens, the small models of data the AI learns from. The aim is to maintain these chips busy on a regular basis, as a result of if the info or the community can’t sustain, every thing slows down.
The AI coaching itself is a cycle: the AI processes tokens in sequence, makes predictions concerning the subsequent one, checks them towards the appropriate solutions and adjusts itself. This repeats trillions of occasions till the system will get higher at no matter it’s being skilled to do. Consider it like an expert soccer staff’s follow. Every GPU is a participant working a drill, the tokens are the performs being executed step-by-step, and the community is the teaching workers, shouting directions and protecting everybody in sync. The staff repeats performs time and again, correcting errors till they’ll execute them completely. By the top, the AI mannequin, just like the staff, has mastered its technique and is able to carry out beneath actual sport situations.
AI infrastructure at frontier scale
Function-built infrastructure is vital to with the ability to energy AI effectively. To compute the token math at this trillion-parameter scale of main AI fashions, the core of the AI datacenter is made up of devoted AI accelerators (corresponding to GPUs) mounted on server boards alongside CPUs, reminiscence and storage. A single server hosts a number of GPU accelerators, linked for high-bandwidth communication. These servers are then put in right into a rack, with top-of-rack (ToR) switches offering low-latency networking between them. Each rack within the datacenter is interconnected, making a tightly coupled cluster. From the surface, this structure seems like many unbiased servers, however at scale it features as a single supercomputer the place tons of of 1000’s of accelerators can practice a single mannequin in parallel.
This datacenter runs a single, large cluster of interconnected NVIDIA GB200 servers and thousands and thousands of compute cores and exabytes of storage, all engineered for essentially the most demanding AI workloads. Azure was the primary cloud supplier to convey on-line the NVIDIA GB200 server, rack and full datacenter clusters. Every rack packs 72 NVIDIA Blackwell GPUs, tied collectively in a single NVLink area that delivers 1.8 terabytes of GPU-to-GPU bandwidth and provides each GPU entry to 14 terabytes of pooled reminiscence. Relatively than behaving like dozens of separate chips, the rack operates as a single, large accelerator, able to processing an astonishing 865,000 tokens per second, the best throughput of any cloud platform accessible at this time. The Norway and UK AI datacenters will use comparable clusters, and reap the benefits of NVIDIAs subsequent AI chip design (GB300) which presents much more pooled reminiscence per rack.
The problem in establishing supercomputing scale, significantly as AI coaching necessities proceed to require breakthrough scales of computing, is getting the networking topology excellent. To make sure low latency communication throughout a number of layers in a cloud surroundings, Microsoft wanted to increase efficiency past a single rack. For the newest NVIDIA GB200 and GB300 deployments globally, on the rack stage these GPUs talk over NVLink and NVSwitch at terabytes per second, collapsing reminiscence and bandwidth boundaries. Then to attach throughout a number of racks right into a pod, Azure makes use of each InfiniBand and Ethernet materials that ship 800 Gbps, in a full fats tree non-blocking structure to make sure that each GPU can discuss to each different GPU at full line fee with out congestion. And throughout the datacenter, a number of pods of racks are interconnected to scale back hop counts and allow tens of 1000’s of GPUs to operate as one global-scale supercomputer.
When specified by a standard datacenter hallway, bodily distance between racks introduces latency into the system. To deal with this, the racks within the Wisconsin AI datacenter are specified by a two-story datacenter configuration, so along with racks networked to adjoining racks, they’re networked to extra racks above or beneath them.
This layered strategy units Azure aside. Microsoft Azure was not simply the primary cloud to convey GB200 on-line at rack and datacenter scale; we’re doing it at large scale with prospects at this time. By co-engineering the total stack with the most effective from our trade companions coupled with our personal purpose-built techniques, Microsoft has constructed essentially the most highly effective, tightly coupled AI supercomputer on this planet, purpose-built for frontier fashions.
Addressing the environmental impression: closed loop liquid cooling at facility scale
Conventional air cooling can’t deal with the density of recent AI {hardware}. Our datacenters use superior liquid cooling techniques — built-in pipes flow into chilly liquid instantly into servers, extracting warmth effectively. The closed-loop recirculation ensures zero water waste, with water solely wanted to replenish as soon as after which it’s frequently reused.
By designing purpose-built AI datacenters, we had been capable of construct liquid cooling infrastructure into the ability on to get us extra rack-density within the datacenter. Fairwater is supported by the second largest water-cooled chiller plant on the planet and can repeatedly flow into water in its closed loop cooling system. The recent water is then piped out to the cooling “fins” on both sides of the datacenter, the place 172 20-foot followers chill and recirculate the water again to the datacenter. This method retains the AI datacenter working effectively, even at peak hundreds.
Over 90% of our datacenter capability makes use of this technique, requiring water solely as soon as throughout development and frequently reusing it with no evaporation losses. The remaining 10% of conventional servers use out of doors air for cooling, switching to water solely through the hottest days, a design that dramatically reduces water utilization in comparison with conventional datacenters.
We’re additionally utilizing liquid cooling to assist AI workloads in lots of our current datacenters; this liquid cooling is completed with Warmth Exchanger Models (HXUs) that additionally function with zero-operational water use.
Storage and compute: Constructed for AI velocity
Trendy datacenters can comprise exabytes of storage and thousands and thousands of CPU compute cores. To assist the AI infrastructure cluster, a wholly separate datacenter infrastructure is required to retailer and course of the info used and generated by the AI cluster. To provide you an instance of the dimensions — the Wisconsin AI datacenter’s storage techniques are 5 soccer fields in size!
We reengineered Azure storage for essentially the most demanding AI workloads, throughout these large datacenter deployments for true supercomputing scale. Every Azure Blob Storage account can maintain over 2 million learn/write transactions per second, and with thousands and thousands of accounts accessible, we will elastically scale to satisfy nearly any knowledge requirement.
Behind this functionality is a basically rearchitected storage basis that aggregates capability and bandwidth throughout 1000’s of storage nodes and tons of of 1000’s of drives. This allows scale to exabyte scale storage, eliminating the necessity for handbook sharding and simplifying operations for even the most important AI and analytics workloads.
Key improvements corresponding to BlobFuse2 ship high-throughput, low-latency entry for GPU node-local coaching, guaranteeing that compute sources are by no means idle and that large AI coaching datasets are at all times accessible when wanted. Multiprotocol assist permits seamless integration with various knowledge pipelines, whereas deep integration with analytics engines and AI instruments accelerates knowledge preparation and deployment.
Computerized scaling dynamically allocates sources as demand grows, mixed with superior safety, resiliency and cost-effective tiered storage, Azure’s storage platform units the tempo for next-generation workloads, delivering the efficiency, scalability and reliability required.
AI WAN: Connecting a number of datacenters for a good bigger AI supercomputer
These new AI datacenters are a part of a worldwide community of Azure AI datacenters, interconnected by way of our Broad Space Community (WAN). This isn’t nearly one constructing, it’s a few distributed, resilient and scalable system that operates as a single, highly effective AI machine. Our AI WAN is constructed with development capabilities in AI-native bandwidth scales to allow large-scale distributed coaching throughout a number of, geographically various Azure areas, thus permitting prospects to harness the ability of a large AI supercomputer.
This can be a elementary shift in how we take into consideration AI supercomputers. As an alternative of being restricted by the partitions of a single facility, we’re constructing a distributed system the place compute, storage and networking sources are seamlessly pooled and orchestrated throughout datacenter areas. This implies higher resiliency, scalability and adaptability for patrons.
Bringing all of it collectively
To satisfy the vital wants of the most important AI challenges, we would have liked to revamp each layer of our cloud infrastructure stack. This isn’t nearly remoted breakthroughs, however composing a number of new approaches throughout silicon, servers, networks and datacenters, resulting in developments the place software program and {hardware} are optimized as one purpose-built system.
Microsoft’s Wisconsin datacenter will play a vital position in the way forward for AI, constructed on actual know-how, actual funding and actual neighborhood impression. As we join this facility with different regional datacenters, and as each layer of our infrastructure is harmonized as a whole system, we’re unleashing a brand new period of cloud-powered intelligence, safe, adaptive and prepared for what’s subsequent.
To study extra about Microsoft’s datacenter improvements, take a look at the digital datacenter tour at datacenters.microsoft.com.
Scott Guthrie is liable for hyperscale cloud computing options and companies together with Azure, Microsoft’s cloud computing platform, generative AI options, knowledge platforms and knowledge and cybersecurity. These platforms and companies assist organizations worldwide resolve pressing challenges and drive long-term transformation.
