CommonAI has selected Kao Data’s Harlow campus to host at least five high-density AI clusters for the UK’s Scaling Inference Lab, Kao Data said in its 21 October 2026 announcement. The agreement gives CommonAI, a non-profit membership organisation, space to test emerging AI infrastructure at rack scale and give UK startups access to the systems.
Key points
- CommonAI plans at least five clusters across two technology suites at Kao Data’s Harlow campus.
- ARIA committed £50 million to the lab, while the UK Government announced more than £20 million in additional funding in June.
- The planned systems combine Nvidia and AMD hardware to test different approaches to AI infrastructure.
ARIA’s £50 million lab programme
The Advanced Research and Invention Agency, known as ARIA, launched the Scaling Inference Lab earlier in 2026 with a £50 million commitment, according to Kao Data. The UK Government announced more than £20 million in additional funding in June for a national capability to develop new AI compute systems. ARIA funds the lab and CommonAI delivers it.
Kao Data describes the lab as a neutral setting for testing emerging technologies at rack scale alongside commercial systems. According to Kao Data, CommonAI is a non-profit membership organisation working on shared engineering for foundational AI technologies. Kao Data says startups will be able to connect their technology to complete systems and run tests against emerging workloads, rather than work with an isolated component.
The lab plans to build a new experimental cluster every six months, Kao Data says. That timetable would give successive technologies a place in the programme beyond the initial Harlow deployment. Kao Data also names Quettaflop AI, Callosum and Oriole Networks as technology partners. The programme is intended to give investors and governments evidence from tests of infrastructure that could be used more widely, according to Kao Data’s description of the lab.
Kao Data assigns two Harlow suites
Under the agreement, CommonAI will place the clusters in two dedicated technology suites at Harlow, Kao Data says. The company develops and operates data centres for AI and advanced computing, with facilities east and west of London. It says its portfolio includes more than 237 MW of IT load across capacity that is operating, under development or planned.
Kao Data says CommonAI chose Harlow for its energy-efficiency credentials, experience with high-density workloads and location between London and Cambridge. David Stringer, Partnerships Lead for the Scaling Inference Lab at CommonAI, said the campus would let the organisation bring startups, investors and policymakers into the lab. Kao Data says the planned clusters will use 100% certified renewable energy.
Spencer Lamb, Kao Data’s chief executive, described the lab as part of demand for UK-based AI infrastructure and resources for startups. Hosting it adds a research programme to a campus that Kao Data says already serves AI and supercomputing workloads. CommonAI is responsible for delivering the lab, while Kao Data supplies the site for the planned deployment.
Nvidia and AMD systems for five clusters
Kao Data says the deployment will include Nvidia HGX B300 systems based on Blackwell Ultra, Nvidia H200 NVL and RTX PRO 6000 Blackwell GPUs, and AMD Instinct MI350X accelerators. The hardware is to sit in servers from Supermicro, MSI, Fujitsu PRIMERGY and Dell PowerEdge. Each cluster is intended to test a distinct way of building AI infrastructure.
The approaches Kao Data identifies include pipeline parallelism, which splits work across processing stages, and heterogeneous orchestration, which coordinates different kinds of computing hardware. The company also lists nanosecond optical circuit switching, an approach to connecting systems using rapidly reconfigured optical links. Testing those approaches beside commercial systems is the purpose of the shared rack-scale environment, Kao Data says.
CommonAI aims to have all five clusters operational by early 2027, Data Centre Magazine reported.
