Neoclouds Becoming AI Capacity Wholesalers
Renen Hallak, CEO of VAST Data, on AI agents creating infinite storage demand
This shows the neoclouds are no longer just overflow vendors for startups, they are becoming a wholesale supply layer for the AI market. When a hyperscaler cannot get enough GPUs, power, or finished cluster capacity online fast enough, it can buy or route demand into a neocloud that already has racks live. That matters for VAST because these operators standardize infrastructure quickly, and once a storage stack is chosen across a fast growing fleet, it can spread across both direct neocloud demand and hyperscaler spillover demand.
-
The practical bottleneck is not just chips. It is power, buildings, networking, and the work of turning thousands of GPUs into usable clusters. Fluidstack moved from a marketplace into private clouds with 100,000 plus GPUs and long dated hosting deals, which is exactly the kind of ready capacity a larger cloud can lean on when its own buildout lags demand.
-
This is already visible in adjacent deals. Lambda signed a multibillion dollar Microsoft agreement to deploy tens of thousands of NVIDIA GPUs, and Google has been using third party partner models for distributed cloud and sovereign offerings. The line between hyperscaler and neocloud is becoming more of a capacity sourcing relationship than a clean competitive split.
-
For VAST, this helps explain why scale customers converge on one modern data layer. Neoclouds need storage that can feed very large GPU clusters without manual data shuffling, and VAST has turned that into major platform deals, including a $1.17B commercial agreement with CoreWeave and broad use across AI focused cloud providers such as Fluidstack.
The market is heading toward fewer, larger neoclouds that look more like infrastructure wholesalers. The winners will secure power, financing, and customer commitments early, then sell capacity both to model builders and to hyperscalers that need faster time to market. That makes the neocloud layer more durable, and makes shared infrastructure suppliers like VAST more deeply embedded in the stack.