QuettaFlow VasEdge

The case for this is not efficiency, it is location. A 12kW cabinet that runs at under 55dB and needs no raised floor can sit in an office, a lab or a plant room, which is where a lot of AI work actually has to happen. The alternative is not a better-cooled datacentre, it is not having a datacentre at all.

Deployment is 400V power and an external dry cooler, and QuettaFlow states four hours to live. The dry cooler is a separate 70kg unit at 1080 x 362 x 720mm, so siting it is part of the survey rather than an afterthought.

It takes Lead Wealth modular servers, where CPU, GPU and storage arrive as separate blocks that combine, so a deployment can start smaller than a fixed chassis would allow and grow without replacing the cabinet. QuettaFlow publishes ready-made configurations around DeepSeek at 32B, 70B and the full 671B, and will build to a specified NVIDIA card instead.

Weight is the number to check against your floor before anything else: 190kg empty and 360kg loaded, in a footprint under a square metre. That is a point load an office floor may not be rated for.