Making one component more efficient does not guarantee that a whole data centre uses less energy. The growing density of AI computing makes the relationships between servers, cooling and power equipment increasingly important.
On 29 September 2026, Yokogawa announced that its Singapore subsidiary had joined the Sustainable Tropical Data Centre Testbed Phase 2.0 consortium, led by the National University of Singapore’s College of Design and Engineering. The initiative is expanding toward a multi-megawatt pilot facility to evaluate operating models for AI data centres in tropical conditions. Yokogawa’s official announcement
This is a research and engineering collaboration. The announcement does not report a completed fleet-wide deployment or a verified percentage saving from Yokogawa’s participation.
Heat and humidity change the engineering problem
NUS identifies high temperatures, humidity, increasing rack density and large cooling demands as central challenges for tropical data centres. Rack density refers to how much equipment and power consumption are concentrated within a server rack. NUS’s testbed mission
That concentration matters because the electricity used by computing equipment ultimately becomes heat that the cooling system must remove. A plan for more processing capacity therefore needs a corresponding plan for moving heat safely.
The testbed’s research areas include direct-to-chip liquid cooling, which takes coolant close to heat-generating components, and immersion cooling, which places suitable equipment in a cooling liquid. It also studies evaporative cooling and digital twins. These are different techniques to evaluate, rather than a claim that one will suit every facility. Research areas listed by NUS
The new contribution is industrial control
Yokogawa brings experience in measurement, automation and energy management. The consortium aims to assess computing, power, cooling, controls and operations as an integrated system, while developing a further research agenda where current commercial solutions fall short. Consortium objectives and Yokogawa’s role
A digital twin is a digital representation used to examine how a physical system behaves. For a data centre, an engineering team might use such a model to compare operating settings before applying changes to the equipment. That is an illustration of the concept, rather than a result already demonstrated by this partnership.
Our interpretation is that a live pilot can expose trade-offs a component specification misses. A setting that reduces one pump’s electricity use could also change temperatures elsewhere. Testing the system together helps establish whether a local gain translates into an overall benefit.
Evidence needs a clearly defined boundary
The useful measurements will include energy and water use at comparable computing loads, equipment temperatures and reliable operation over time. Efficiency claims are easier to assess when the test conditions and system boundary are disclosed.
The announcement also refers to Singapore’s plans for up to 700 MW of next-generation data-centre capacity on Jurong Island. That wider development target is separate from the testbed’s pilot scale. Pilot and national infrastructure context
The project’s contribution will depend on what the experiments show. Validated operating data could help tropical facilities make better choices about cooling and controls as AI workloads grow.
Featured image: Representative server equipment; not the NUS testbed or a Yokogawa installation. Photo: Tyler / Unsplash.
Leave a Reply