Data-center capacity
We source and contract power and space for your GPU systems, in blocks from 250 kW to 1 MW.
Capacity in fixed blocks
Each block comes with the space, cooling and term to match.
We source and contract GPU colocation in three block sizes, 250 kW, 500 kW and 1 MW, each matched to the systems you run.
- CoolingLiquid-cooled or rear-door cooled.
- Term24 months as standard.
- PricingPer kW-month.
A kW-month is one kilowatt of a block's power for one month. A 500 kW block on a 24-month term is 12,000 kW-months.
Ask for a blockSized to your systems
Each block is matched to the GPU systems it runs, by power draw and cooling grade.
Power draw
The block covers what your systems draw and the cooling they need, rack by rack.
Your systems
The power and cooling each GPU system needs are on our GPU systems page.
GPU systems
What fits in each block
How many whole GPU systems each block powers, worked out from the manufacturers' published power figures.
| System | Power, maximum | Per 250 kW block | Per 500 kW block | Per 1 MW block |
|---|---|---|---|---|
| HGX B300 8-GPU system | 15 kW | 16 | 33 | 66 |
| GB300 NVL72 rack | 142 kW | 1 | 3 | 7 |
The counts are illustrative: whole systems at the manufacturer's maximum power, before the switches, storage and other equipment a block also powers
Power, HGX B300 8-GPU system: maximum for NVIDIA's DGX B300 system. Source: NVIDIA, read 6 Nov 2026
Power, GB300 NVL72 rack: maximum for a full rack. Source: NVIDIA, read 6 Nov 2026
How the blocks are cooled
Every watt a rack draws leaves it as heat, so the cooling sets how much power each rack can use.
Air cooling
Fans pull room air through each server, and the room's own cooling carries the heat away.
Rear-door cooling
A heat exchanger in the rack's rear door, with liquid running through it, takes the heat out of the servers' exhaust air before it enters the room. The servers inside stay air-cooled.
Liquid cooling
Liquid carries the heat out of the servers themselves and away from the rack. Some systems, such as NVL72 racks, are built for liquid cooling.
Density is counted in kW per rack. NVIDIA gives a maximum of 142 kW for a full GB300 NVL72 rack, and 15 kW for each 8-GPU B300 system a rack holds.
The blocks we contract are liquid-cooled or rear-door cooled, for high-density racks of GPU systems. Systems built for liquid cooling, such as NVL72 racks, need liquid-cooled colocation. Each quote states the block's cooling.
What a capacity quote states
Every capacity quote states the same five points.
- Block250 kW, 500 kW or 1 MW of power, with the space to hold your systems.
- CoolingLiquid cooling or rear-door cooling, matched to the systems the block runs.
- Term24 months as standard.
- Price per kW-monthThe price of each kilowatt of the block for each month of the term.
- Start monthThe month the power and space are ready for your systems.
Reserve now or ahead
Contract capacity for today, or reserve it ahead of need so power and space are ready when the hardware arrives.
- NowPower and space for GPU systems you already hold or have on order.
- Ahead of needPower and space reserved for a named start month, before the hardware arrives.
- With a forward agreementPair a capacity agreement with a forward purchase agreement to line up power and space with the delivery month.
- Finance itGPU financing for the systems the block runs, arranged with lending and leasing partners.
How capacity comes together
Four steps, from your systems to contracted power and space.
Request
Tell us the systems, the power they draw, the cooling they need, the start month and the location.
Source
We find power and space that fit the systems, the cooling grade and the start month.
Contract
The agreement sets the block, the cooling, the term and the price per kW-month.
Move in
Your systems arrive in the agreed month and run on the contracted power.
Common questions
What is a kW-month?
One kilowatt of a block's power for one month. Capacity is priced per kW-month, so a 500 kW block on a 24-month term is 12,000 kW-months.
How much power does a GPU rack draw?
It depends on the systems it holds. NVIDIA gives a maximum of 142 kW for a full GB300 NVL72 rack and 15 kW for an 8-GPU B300 system, so a rack of 8-GPU systems draws up to 15 kW for each system it holds, plus its switches and storage. What fits in each block gives the source of each figure.
Which cooling do the blocks use?
Liquid cooling or rear-door cooling, matched to the systems each block runs. Systems built for liquid cooling, such as NVL72 racks, take a liquid-cooled block. Each quote states the block's cooling.
Research
The newest articles from our Insights library, each with a full research report and its sources.
- What causes data-center outages: surveys and incident reviewsUptime's 2025 survey puts power behind 45% of impactful outages. Published incident reviews show how faults in power, cooling and water spread, and what set recovery time.Data-center capacity7 min readPDF report
- Colocation pricing per kW: how the monthly bill is builtWholesale colocation is quoted per kW per month. Lease structure, power pass-through, escalators, one-time fees and the billing start date set the bill, as public filings show.Data-center capacity7 min readPDF report
- Data-center Tier levels: what Tier III and IV coverUptime Institute's four data center Tier levels set maintenance and fault-tolerance criteria with no uptime percentage attached, and NVIDIA says dual feeds give a DGX B200 rack only N.Data-center capacity7 min readPDF report
Notices
Forward agreements. Forward agreements are bilateral purchase agreements for physical delivery, negotiated per order and not transferable. Rillor does not operate an exchange or offer futures or other derivatives.