GPU containers.
Built for AI.

From energized power to live compute.
Ready for service in weeks.

QIT M1 is a pre-engineered, chip-agnostic 1 MW compute module. Rack NVIDIA B200, B300 or Vera Rubin, or AMD Helios, at sites with available power. Modular data centers for AI inference, training and your next compute workload.

Discuss a project
Concept illustration of a QIT M1 GPU container with server racks and liquid-cooling pipes
QIT M1
Chip-agnostic 1 MW IT module, phased to 5–10 MW
Cooling
Direct liquid cooling, zero water use
Delivery
Ready for service in weeks on an energized site

Start with power.
Build the
compute.

The starting point is a site with power already available. QIT develops modular AI data centers around that capacity, bringing liquid-cooled GPU containers, networking and supporting infrastructure into one coordinated deployment.

We assess the site and workload together: usable electrical capacity, connectivity, cooling and room to expand. Module integration and site preparation move in parallel, with the interfaces and responsibilities defined from the start.

Explore the platform

QIT M1 modules are integrated and tested in the factory, then set and connected on site in about 15 hours. No construction project on site: once power and the pad are ready, compute goes live.

For inference providers, AI teams and owners of powered sites, the goal is the same: a practical path to live compute. GPU selection, operating scope and acceptance criteria are agreed for each project.

Start a conversation
01

One partner

QIT takes responsibility for the whole stack: GPU racks and networking, liquid cooling and heat rejection, power distribution and site interfaces. One team from scoping to go-live.

02

Plug & play commissioning

Every container is integrated, tested and commissioned in the factory, then shipped ready to set on a prepared pad. On an energized site, the date that matters is ready for service, measured in weeks.

03

Scalable architecture

Each QIT M1 delivers 1 MW of IT capacity and is chip-agnostic: rack what you want. Start with a single module and add more as site power allows, up to a full campus.

Conceptual plan of GPU container modules, liquid cooling equipment and site power connections
GPU platform, rack count, fabric and storage are defined around your workload. GPU container guide
  1. Step 01

    Scope & site

    Assess available power and connectivity, then agree the workload, configuration, interfaces and responsibilities.

  2. Step 02

    Factory build

    Integrate and test the container in the factory while the site team prepares power, cooling and connectivity.

  3. Step 03

    Deliver

    Ship the tested container to the prepared site.

  4. Weeks

    Install & go live

    Set and connect in about 15 hours, then commission to ready for service.

The on-site work starts when the container arrives at a prepared site with power, pad and connectivity in place.

Cutaway drawing of the QIT Pod 001 container with six customer GPU racks, coolant distribution unit and overhead coolant manifold
IT load
6 × NVIDIA GB300 NVL72, 810 kW TDP / 930 kW peak
Facility
1.15–1.25 MW, 7th rack position reserved
Power
Existing host transformer reused, 415Y/240 V, 2,500 A
Cooling
1 MW liquid-to-liquid CDU, adiabatic dry coolers

A liquid-cooled 40 ft hosting container for customer-owned racks, placed on an operating bitcoin-mining site and fed from power already on the property. Preliminary design, issued for study.

What do you mean by a GPU container?

A physical modular data center in a shipping-container-style enclosure, designed around GPU compute and its supporting infrastructure. QIT focuses on liquid-cooled GPU deployments for AI workloads.

How fast can QIT install a container?

On a site whose utility connection is already energized, a container is set and connected in about 15 hours and reaches ready for service within weeks. Transformer, switchgear and utility lead times set the date where they are still open. Each container is integrated and tested in the factory, then set, connected and commissioned at a prepared site with power, pad and connectivity in place.

Does QIT M1 support NVIDIA Vera Rubin?

Yes. QIT M1 supports NVIDIA Vera Rubin, with direct liquid cooling to every GPU and CPU. Power distribution and cooling are sized to the Vera Rubin rack envelope for each project, alongside B200, B300 and AMD Helios.

Which GPUs and how much power can a container support?

Each 20 ft QIT M1 delivers 1 MW of IT capacity. QIT M1 is chip-agnostic: rack what you want, including NVIDIA B200, B300 and Vera Rubin, and AMD Helios, with direct liquid cooling on every GPU and CPU.

What does the site need to provide?

A prepared location, appropriate power and network connectivity, cooling interfaces and delivery access. The split between QIT’s scope and site infrastructure is agreed during planning.

Is QIT an NVIDIA DSX POD or Crusoe partner?

QIT is independent. DSX POD and Crusoe Spark are useful industry reference points, but this site does not claim an NVIDIA or Crusoe partnership, certification or validated design.

How do I get a project proposal?

Prepare a brief with your location, available power, workload and desired deployment date, then email it to ismail@qit.energy. Our project brief form helps organize these inputs.

Let’s put your
power to work.

Start with your site, your workload and the compute you need.

Discuss a project
ismail@qit.energy