At a glance
- CommonAI will deploy high-density, GPU-accelerated systems in two dedicated technology suites, Kao Data said this week.
- For data center underwriters, the agreement as reported carries no contract value and no capacity figure for the two suites.
- The five GPU systems mix Nvidia HGX B300, H200 NVL and RTX PRO 6000 Blackwell GPUs with AMD Instinct MI350X accelerators, housed in Supermicro, MSI, Fujitsu Primergy and Dell PowerEdge servers.
Kao Data will host a UK government-funded AI inference lab at its Harlow campus, with CommonAI deploying five GPU systems in two dedicated technology suites.
CommonAI will deploy high-density, GPU-accelerated systems in two dedicated technology suites, Kao Data said this week. The Scaling Inference Lab is funded by the UK's Advanced Research & Invention Agency (ARIA) and delivered by CommonAI, according to Data Center Dynamics, and is intended as a neutral environment for testing emerging AI against commercial systems.
For data center underwriters, the agreement as reported carries no contract value and no capacity figure for the two suites. ARIA funds the lab, CommonAI runs it, and Kao supplies the site and security; lease term and megawatt commitment would show whether this is a demonstration tenancy or anchor demand.
Five GPU systems, four server vendors
The five GPU systems mix Nvidia HGX B300, H200 NVL and RTX PRO 6000 Blackwell GPUs with AMD Instinct MI350X accelerators, housed in Supermicro, MSI, Fujitsu Primergy and Dell PowerEdge servers. Each deployment tests a different approach to AI infrastructure: pipeline parallelism, heterogeneous orchestration and nanosecond optical circuit switching.
Harlow opened in 2018 and was originally planned for four 10MW buildings. Two are running, KLON-01 and KLON-02, 10MW each. Kao announced KLON-03, a 17.6MW facility, in March 2025 and plans for two more buildings in September 2025, taking the campus to five.
The campus sits between London and Cambridge and already houses Nvidia's Cambridge-1 supercomputer in Kao's original Harlow building; Nebius is also taking capacity there.
All five GPU systems are due to be running by early 2027.
Reported by the Data centers & fiber desk from Data Center Dynamics. Published . Corrections