AWS, CoreWeave top AI infra shortlist

AI infrastructure is turning into its own buying category. Not just “which cloud do we already use,” but “where do we actually run and serve giant models without blowing up cost, latency, or timelines.” That’s why this Constellation ShortList matters. In Q1 2026, Constellation published a dedicated “AI Factory Providers” list, and AWS and CoreWeave landed in the top group alongside a mix of hyperscalers, Nvidia-linked offerings, and newer AI-native clouds. (constellationr.com) ### What is an “AI factory” here? Basically, it means the full stack for industrial-scale AI — GPU clusters, networking, storage, orchestration, and the operational plumbing to train, fine-tune, and serve models in production. This is not the same thing as a generic cloud VM catalog. Constellation split(constellationr.com)hich is a clue that buyers are now distinguishing between broad enterprise platforms and specialist GPU clouds. (constellationr.com) ### What actually changed? The new part is the formal shortlist itself. Constellation’s Q1 2026 update added or refreshed 83 categories across two waves, and “AI Factory Providers” appeared as one of the categories tracking where enterprise demand is consolidating. The listed vendors are AWS, CoreWeave, (constellationr.com)loud Infrastructure. (constellationr.com) ### Why are AWS and CoreWeave the eye-catching names? Because they represent the two poles of the market. AWS is the classic hyperscaler play — huge ecosystem, lots of adjacent services, custom silicon, and the ability to bundle AI into broader enterprise infrastructure. (constellationr.com) AI teams rather than general IT buyers. (constellationr.com) ### Why isn’t this just “AWS wins”? Because AI teams care about different bottlenecks than normal cloud buyers. A company training or serving a frontier-scale model may care more about getting the right GPUs now than about having the deepest menu of databases and middleware. That’s wh(constellationr.com)“AI Clouds with Compute Roots” list leans heavily in that direction, with CoreWeave, Crusoe, Lambda Labs, Nebius, RunPod, and Together.ai. (constellationr.com) ### So is price the whole story? No — but it’s a big part of it. Third-party pricing trackers and comparisons consistently frame CoreWeave as cheaper on equivalent Nvidia GPU capacity, sometimes by a lot, while AWS trades that off for broader integration, procurement comfort, and enterprise features. You should treat exact(constellationr.com)lity change fast. But the shape of the tradeoff is real. (deploybase.ai) ### What is AWS doing to answer that? AWS is trying to meet the specialist-cloud challenge from both ends. It keeps offering newer Nvidia-based instances, but it is also pushing “AI Factories” as a named product and leaning hard on Trainium custom silicon to lower cost for some workloads. That matters because the hyperscaler answer to GPU scarc(deploybase.ai)ips where we can, and package the whole stack as an AI factory.” (constellationr.com) ### What does this say about the market now? Turns out the market is no longer one race. It’s two. One race is breadth — who can give enterprises the safest, most integrated place to build AI. The other is raw AI throughput — who can deliver the best GPU access, utilization, and economics right now. AWS shows up as the broad winner candidate. CoreWeave shows up as the specialist benchmark. (constellationr.com) ### Bottom line? This shortlist doesn’t prove AWS or CoreWeave is “best” in the abstract. It shows that AI infrastructure buying has matured enough that buyers are now making an explicit choice between two very different answers — hyperscaler convenience or specialist GPU focus. And in 2026, both answers are winning. (constellationr.com)