Liquid-cooled GPU capacity in Europe, without the grid queue

We install 500 kW liquid-cooled pods in concrete buildings that are already built. A pod goes live 10 to 15 days after it reaches the site.

Three black BlueHarbor server racks with glass doors and blue status lights, with the BlueHarbor logo on the side panel. Render
5 MWIT capacity at Campus I
500 kWPer pod
150 kWPer rack, liquid-cooled
≤ 1.20PUE

Campus I is two concrete frames in Portugal. The concrete is done. What is left is to bring in ten pods, 5 MW in total.

An empty three-storey reinforced concrete frame under a blue sky.
Today. Photo.
The same concrete frame with white container pods on each floor and steel exhaust stacks.
With pods installed. Render.

A new data centre in Europe's main hubs waits 7 to 10 years for a grid connection. Campus I does not need a new building.

012345678910 Years from project start New build BUILDING AND GRID QUEUE, NO CAPACITY GRID CONNECTION ARRIVES, YEAR 7 TO 10 Campus I indicative 1 MW, 3 MW, THEN 5 MW BY QUARTER 5
New buildCampus I
BuildingTo design, permit and constructTwo concrete frames, already built
Work on siteFull constructionSlab strengthening, power, pipework, security
First capacityAfter the building and a 7 to 10 year grid queueFirst pod, 10 to 15 days after delivery

Portugal is where capacity can land faster.

Campus I puts 5 MW of liquid-cooled compute into concrete frames that are already standing.

One pod is a 40-ft container with 500 kW of IT capacity. Racks, power and cooling leave the factory as one unit.

Cutaway of a 40-ft container with a row of racks, overhead coolant pipes, a coolant distribution unit and a hose manifold on the open door. Render, pod with one side removed
  1. 1

    Up to 150 kW per rack

    Cold plates on the chips carry the heat away in liquid. Air cooling stops long before that density.

  2. 2

    PUE of 1.20 or lower

    About 40% less cooling energy than an air-cooled hall. You pay for electricity at cost, so the saving is yours.

  3. 3

    Pods can move

    A pod is equipment, not part of the building. It can go to another site.

  4. 4

    Run by engineers who have done it at hyperscale

    Senior operators from Alibaba Cloud's data-centre organisation specify the pods and run them 24/7. Meet the team

The aisle inside a pod, with rack doors on both sides and blue coolant pipes along the ceiling.
Inside the podOne row of racks, with the coolant lines overhead. Render.
Close-up of metal cold plates on a server board, with blue coolant hoses on quick-connect fittings.
Cold plates on the chipsLiquid carries the heat out of the rack. Render.
The rear of a rack, with two steel manifolds and rows of blue coolant hoses.
Rack manifoldSupply and return for every server. Render.

Capacity goes live one pod at a time.

Dark 40-ft pods in a factory hall, with racks and cooling plant being fitted inside.
10 to 16 weeks in the factoryRacks, power and cooling go in before the pod ships. Render.
A crane lifts a pod into a bay of a three-storey concrete frame at dusk, next to pods that are already lit.
7 to 10 days on siteThe pod is lifted into the existing frame, then commissioned in 3 to 5 days. Render.

Time per pod

On site, a pod takes 10 to 15 days from delivery to service.

Grow in 500 kW steps

1 MW, then 3 MW, then 5 MW. The first pods earn while the next ones go in.

Phase 1
1 MW
Phase 2
3 MW
Phase 3
5 MW

About 40% less cooling energy

Cooling energy. Air-cooled hall = 100.

Air-cooled100
BlueHarbor~60

One pod design for every site

The site search for Campus II is under way.

Campus IPortugal, 10 pods
Campus IISite search

Use the capacity, or own part of it.

AI and compute teams

Reserve whole 500 kW pods at Campus I. Our engineers run them 24/7.

Rack density
Up to 150 kW
Capacity unit
500 kW pod
Electricity
At cost
Operations
24/7 on site

Investors

BlueHarbor is raising equity from financial and industrial investors. After a first call, we send the Campus I investor materials.

Three capabilities that rarely sit in one place. A team that has built AI businesses, engineers who have run hyperscale data centres, and the local access to land them in Portugal.

Sweden

Management

Founders who have built AI and fintech companies and raised money from institutional investors.

  • Institutional fundraising
  • Hardware sales
  • Nordic governance and reporting
Hyperscale

Build and operations

Senior engineers from Alibaba Cloud's data-centre organisation. They specify every BlueHarbor pod and run it 24/7.

  • Nvidia certified
  • Liquid cooling to 150 kW per rack
  • Commissioning at scale
Portugal

Local access

The contacts that move grid access, permits and incentives in Portugal.

  • Top-tier Portuguese law firm
  • Co-investor network
  • National and municipal government

Request capacity at Campus I

Tell us how much capacity you need and when. Investors can write to the capital team directly.