Constellation lists top AI factory clouds
- Constellation Research added a new “AI Factory Providers” ShortList on February 25, naming AWS, CoreWeave, Crusoe, Dell with Nvidia, Google, Microsoft, Oracle, IBM, Nebius, and Nvidia. - The screen was unusually concrete: vendors needed 5-plus AI factory locations, 4 machine sizes, 200 paying customers, three continents, and generative AI support. - It matters because AI buying is shifting from generic cloud to power-dense, sovereign, factory-style capacity where location and expansion now matter.
AI cloud buying just got a little more specific. Constellation Research used its Q1 2026 ShortList refresh to add a brand-new category called “AI Factory Providers,” and the names on it tell you where enterprise demand is heading. This is not a general cloud ranking. It is a list for companies that need dense AI infrastructure — the kind built for training models, serving inference at scale, and handling the ugly physical realities behind both. The bigger story is that “AI factory” is turning into a real buying category, not just Nvidia marketing. ### What is an AI factory here? Constellation is basically using the term to mean a cloud or infrastructure provider that can run generative AI workloads at industrial scale, across multiple locations, with enough operational maturity that a big customer could actually buy from it. The new shortlist includes Amazon Web Services, CoreWeave, Crusoe, Dell AI Factory with Nvidia, Google Cloud Platform, IBM Cloud, many newer GPU specialists and infrastructure stacks built around Nvidia. ### Why is this different from plain old cloud? Because the bottleneck is no longer just “who has servers.” It is who has the right kind of servers, in enough places, with enough scale, and with a roadmap that keeps up with AI demand. Constellation’s threshold for this list required availability on three continents, more than 200 paying customers, more than five AI factory locations with plans to keep adding site certifications. That is a much narrower filter than a generic IaaS bake-off. ### Why are specialists showing up next to AWS and Azure? Because AI has created room for clouds that started with compute first and built outward later. Constellation has a separate ShortList called “AI Clouds with Compute Roots,” published the same day, and that one includes CoreWeave, Crusoe, Lambda Labs, Nebius, RunPod, and Together.ai. The criteria there lean even harder into AI-era expansion — 1,000-plus model-provider partnerships, and the ability to run dedicated government workloads on three continents. Basically, the market now has two lanes: old cloud giants adapting to AI, and AI-native clouds growing into full platforms. ### Why does geography keep coming up? Because AI capacity is no longer useful if it is in the wrong country. Constellation keeps tying these lists to sovereign cloud demand — customers increasingly want workloads kept inside national borders, sometimes with local operational control. That pushes AI infrastructure from a pure performance question into a location and compliance question. The shortlist language even frames this as a race for footprint, resilience, and enough sites to support high availability. ### So what changed in February? The category itself. In Constellation’s February 25 Q1 2026 release, “AI Factory Providers” appeared as a new ShortList alongside other fresh categories like AI Frameworks and Sovereign Clouds. That is the signal. Analysts are carving AI infrastructure into more precise buckets because enterprise buyers are no longer shopping for vague “AI capability.” They are shopping for fit — training-heavy, inference-heavy, sovereign, secure, expandable. ### What does the vendor list say about the market? It says scale still wins, but specialization now counts. AWS, Google, Microsoft, and Oracle bring reach and enterprise trust. CoreWeave, Crusoe, and Nebius bring AI-first focus. Dell with Nvidia shows the on-prem and hybrid angle is real, not a side quest. Nvidia appearing directly on the list is maybe the clearest tell of all — the stack is getting vertically integrated enough that buyers are evaluating “factory” capability, not just renting commodity compute. ### What should buyers take from this? If you are buying AI capacity in 2026, the question is not just price per GPU hour. It is whether a provider can keep expanding, place workloads where you need them, offer enough instance variety, and survive the next few quarters of capex intensity. That is what Constellation’s criteria are really screening for. “AI factory” now means a specific class of infrastructure provider with scale, geography, and operational depth — and that is becoming how serious AI buyers sort the market.
