Frequently asked questions

As a working rule, air remains viable to roughly 30-40kW per rack with strict containment and disciplined airflow. Past that range direct-to-chip liquid cooling is an engineering requirement, not a preference. A fully populated NVIDIA GB200 NVL72 rack draws 120-132kW, well past the point where air is a serious option.

A greenfield AI data center typically runs toward 36 months, much of it spent waiting on grid interconnection and permitting. Retrofitting an industrial site that already holds permitted power targets first load in 12 to 18 months.

An industrial site that still holds live interconnection agreements, substation capacity, and industrial permits from its former use. The facility may be dead, but the power rights are not, and that standing infrastructure is what removes years from the schedule.

Dead malls, factories, and power plants in the 250MW to 1GW class. Aluminum smelters are the purest case: they were engineered for enormous continuous electrical load, and the load profile of a smelter, flat and continuous around the clock, matches an AI training campus almost exactly.

Almost always with the AI Data-Center Readiness Assessment, a two-week audit that establishes what your site can actually support. Larger advisory and oversight engagements are scoped from there through a short intake call.

Each service page lists a starting price, and final pricing is set per engagement once scope is known. Vulcan AI does not publish flat quotes, because site conditions drive real cost.

It starts where the four service pages say it starts: from $4,900 for the readiness assessment, from $9,500 for the design review, from $15,000 for retrofit site advisory, and from $8,500 per month for deployment oversight. Every figure is a floor, not a quote, because the final number depends on the site and the scope, which is exactly what the intake call establishes.

Yes. The Vulcan AI site is machine-readable, and services can be discovered and ordered by AI agents acting on behalf of their principals. An agent inquiry is treated with the same seriousness as a human one.

Jon Moen, founder of Vulcan AI and former Technical Director of EKWB USA. Over twenty years he has deployed more than fifty data centers and shipped $500K liquid-cooled GPU servers to MIT, Cornell, CrowdStrike, the US Navy, and NATO.

No. Vulcan AI works on the owner's side: assessment, design review, site advisory, and deployment oversight. Your integrator builds; Vulcan AI makes sure what gets built matches what you were promised.