Tier III certified facilities·NVIDIA GB300 NVL72 racks allocated

AI Factory operator: whole NVIDIA GB300 NVL72 racks, end-to-end operation

An AI Factory operator with an allocation of 350+ NVIDIA GB300 racks

CambridgeNexus owns and runs bare-metal NVIDIA GB300 NVL72 racks and B300 systems in Tier III facilities in the United States and Taiwan. Leased whole, single tenant, operational within 75-90 days of contract.

Reserve capacity See how we work

Built and delivered with

An AI Factory operator, end to end

CambridgeNexus runs all seven operating layers, from power to workload planning, in Tier III data centers in the United States and Taiwan. Allocation of 350+ NVIDIA GB300 NVL72 racks secured. Terms from 6 months to 5+ years, first workload in 75-90 days from signing.

What you can lease from us

Whole racks or whole servers, single tenant, operated end to end. Pick a size and a term.

NVIDIA B300 Systems

Air-cooled 8-GPU systems. Installed within 75-90 days of contract. For inference and fine-tuning.

GB300 NVL72 Rack Leasing

72 Blackwell Ultra GPUs per rack, liquid-cooled, single tenant. Operational in 75-90 days from contract.

NVIDIA Vera Rubin, allocation secured

CambridgeNexus already holds a Vera Rubin allocation. Racks expected to be available to lease in mid-2027; reservations open now.

Seven layers. One operating model

CambridgeNexus operates the entire stack, from the power feeding the rack to the workload running on it.

See how a deployment runs

Power

Already contracted at every site. Your rack is allocated against power we hold.
Cooling

Liquid or air cooling by configuration: GB300 NVL72 runs liquid-cooled, B300 runs air-cooled. Engineered and operated by us.
Networking

InfiniBand or RoCE fabric for large-scale AI. Up to 400 Gbps external connectivity.
Compute

GB300 NVL72 racks with 72 Blackwell Ultra GPUs or B300 systems. Single tenant.
Orchestration

Scheduling and cluster management configured to your stack.
Compliance

Export-control screening, KYC and data residency fixed in the contract.
Workload planning

We size the deployment around your training or inference run.

Multiple Tier III AI data centers across the United States. CambridgeNexus selects the site to match your latency, compliance and workload requirements.

Tier III certified facilities in the United States and Taiwan

Talk to us about location requirements

Why teams choose CambridgeNexus

Bare-metal NVIDIA GB300 racks, one operator, a delivery schedule instead of a construction project.

See the full comparison

Bare-metal racks

Bare-metal, single-tenant GB300 NVL72 racks, owned and run by CambridgeNexus. No shared tenancy, no fractional GPUs.

One operator, 7 layers

Power, cooling, networking, compute, orchestration, compliance, and workload planning in one contract.

75-90 days vs an industry norm of 6 months

Facilities are built and racks are installed, so your first workload runs 75-90 (or fewer) days from signature. The industry norm is about 6 months.

Built for the workloads you run

Capacity sized around your program,
with the compliance documentation regulated buyers need.

AI Video Generation

Dedicated racks for video-generation model training and inference.

Finance & Insurance

Risk modeling and quant workloads with documented residency and controls.

Biotech & MedTech

Drug-discovery and imaging models on compliant, single-tenant capacity.

Defense & Public Sector

Compute for regulated programs, with export-control screening built in.

Train Your Own Model

Whole racks for foundation-model and fine-tuning runs, on hardware only you use.

Built for the people who review the contract

Compliance is one of the seven layers we operate, built into every deployment from day one.

Read our compliance and security commitments

Export-control screening

Every deployment passes KYC and technical qualification before capacity is confirmed.

Data residency by site

We propose the installation site, US or Taiwan, based on your workload, compliance, and latency, fixed in the contract.

Accountable operations

A named escalation path and one owner for your deployment, from installation to production.

FAQ

What does CambridgeNexus actually do?

CambridgeNexus is an AI Factory operator. It leases bare-metal, single-tenant NVIDIA GB300 NVL72 racks and B300 systems from Tier III data centers in the United States and Taiwan, and operates every layer around them: power, cooling, networking, compute, orchestration, compliance and workload planning. See the seven operating layers

Does CambridgeNexus sell fractional GPUs or cloud instances?

No. CambridgeNexus leases whole, single-tenant hardware: NVIDIA GB300 NVL72 racks and NVIDIA B300 systems, never sliced GPU nodes or shared instances. We may explore additional models in the future.

Are CambridgeNexus data centers already operating?

Yes. CambridgeNexus runs Tier III facilities in the United States and Taiwan, with GB300 racks installed and operational. Technical detail on any site is available on request. How a CambridgeNexus rack is commissioned and run

Who is CambridgeNexus, and who leads the company?

CambridgeNexus, Inc. is an AI Factory operator headquartered at 56 Broad Street, Boston, Massachusetts, led by CEO Ken Huang and an international team across the US, Europe and Asia. Full company details are on the about page.

How fast can CambridgeNexus deploy?

75-90 days from contract to installation and acceptance. Or even faster, depending on rack availability. The clock runs from the contract date. See the GB300 deployment timeline

How do I get pricing from CambridgeNexus?

Price depends on rack count, booking timeline and downpayment, so a list price would be misleading. Send us those three details, and we'll return a quote for your configuration. How CambridgeNexus pricing works