Refreshed weekly · Strictly confidential
Build-to-suit GPU capacity.
Dedicated bare metal B300, HGX B300 and GB300 NVL72 clusters across US, Nordic and EU sites. Some builds are already in motion with committed service dates; the rest are specified around your requirement. We hold no inventory and take no position in the outcome.
End users only. Builds are shown and held for the named ultimate offtaker, not for intermediaries. Full platform coverage and scale bands are on the GPU clusters & GPUaaS desk.
Six answers. Then the builds.
Builds are not published — they move weekly and the parties behind them expect discretion. Answer these and the matching ones appear as you go. Exact rates, sites and confirmed dates come after you submit.
Bare metal, built to your specification.
Dedicated bare metal on a fixed term — physical GPU servers to a single offtaker, not managed cloud sold by the node-hour. US GPU capacity for domestic latency or residency requirements; Nordic and EU builds where the power basis wins.
- —In motion
Builds with committed service dates, marked in the matcher. Faster to stand up, and less room to change the specification.
- —Specify to order
Platform, scale, density and cooling set around your workload. From Q2 2027, and later for anything large.
- —Term sets the rate
Five years prices materially below three on the same hardware. That is supplier funding cost, not a discount.
- —Deposit is the entry condition
30% at three years, 20–25% at five. Nothing is held without one.
Bare metal is one route, not the only one.
Builds and allocation surface continuously and rarely reach an open market. Send the requirement anyway.
Where allocation comes from
Cancelled and restructured orders, incoming stock needing a discreet home, partner positions across regions, and redeployment before end of life.
Open the mandate → GPUaaSReserved capacity instead
Where a dedicated position is heavier than you need — reserved and committed-spend agreements, negotiated on the terms that actually decide the cost.
Open the mandate → DevelopmentPowered shell & build-to-suit
Where the cluster is purpose-built — site selection, power in-service dates contracted separately, liquidated damages and expansion rights.
Open the mandate → CapacityBring your own hardware
If you hold GPUs and need somewhere to land them — AI-ready colocation up to 150 kW per rack, sourced and negotiated globally.
Open the mandate →Asked about the board.
01Why are rates, sites and dates not published?
They are commercially sensitive to the parties behind the builds, and publishing them would anchor a market that moves weekly. Indicative pricing sits in a $3–5 per GPU-hour band, set primarily by term length. Our partners return exact figures within one business day of a qualified enquiry.
02Do you own or resell any of this?
No. We hold no inventory and we are not the seller. You contract directly with the supplier, and we negotiate alongside you.
03Is a deposit really required?
Yes, on every build — 30% at three years, 20–25% at five, because a longer commitment de-risks the supplier. Nothing is held without one, so the matcher will not show builds to a party who cannot place one.
04What does "in motion" mean?
The build has a committed service date and is already underway. Faster to stand up than an open specification, with correspondingly less room to change the configuration.
Rates and sites within one business day.
Six answers gets you the matching builds. Or book thirty minutes and we will talk through them directly.