Johnson Controls is joining Singapore’s Sustainable Tropical Data Centre Testbed 2.0 (STDCT 2.0), a National University of Singapore-led initiative designed to test AI-ready data center technologies under the heat and humidity of a tropical operating environment. The collaboration puts thermal management, smart controls and heat-recovery technologies into a broader effort to increase computing capacity without proportionally increasing power and water consumption.
Artificial intelligence is creating a data center infrastructure problem that cannot be solved simply by adding more servers. As AI accelerators push rack densities higher, operators increasingly need to rethink how facilities consume electricity, reject heat and manage scarce resources. Singapore is turning that challenge into a live engineering experiment.
Johnson Controls has joined the Sustainable Tropical Data Centre Testbed 2.0, hosted by the National University of Singapore’s College of Design and Engineering (NUS CDE) on Jurong Island. The program is intended to evaluate next-generation data center technologies under real tropical conditions, where high temperatures and humidity make cooling more demanding than in many conventional data center environments.
The timing reflects a broader shift in AI infrastructure. Gartner forecasts that global data center electricity consumption will reach 565 TWh in 2026, up 26% year over year. AI-optimized servers are expected to account for 31% of data center power consumption this year, with their electricity use projected to surpass conventional servers in 2027.
The International Energy Agency similarly expects worldwide data center electricity consumption to more than double to roughly 945 TWh by 2030, with AI a major driver of the increase.
That makes cooling and power efficiency increasingly central to the economics of AI infrastructure, rather than secondary facilities concerns.
Johnson Controls says its role in STDCT 2.0 will focus on thermal management, controls, engineering and operational expertise. The company is positioning its technology around an integrated approach in which cooling, power and controls are optimized together rather than treated as separate systems.
One company modeling exercise illustrates the potential scale. Johnson Controls says an integrated thermal-management design could reduce non-IT energy consumption by up to 50% compared with conventional cooling approaches. The company also describes a 1-gigawatt data center reference design using an absorption chiller to capture waste heat from on-site power generation. According to Johnson Controls, the design could convert energy otherwise lost as waste heat into cooling and potentially support up to 97 MW of additional AI computing capacity while reducing cooling-related electrical demand by as much as 44%.
Those figures are company modeling claims rather than independently validated results from STDCT 2.0, an important distinction as the technology moves into a live test environment.
The testbed itself is designed to address precisely this type of validation challenge. NUS describes STDCT 2.0 as a multi-megawatt, AI-ready platform for testing low-carbon data center technologies under tropical operating conditions. Research areas include liquid cooling, direct-to-chip and immersion cooling, enhanced evaporative cooling and digital monitoring and control.
The initiative is also becoming a broader industry platform. NUS has previously announced collaborations involving Schneider Electric, Sembcorp and Eaton, while EdgeConneX joined the STDCT 2.0 consortium as an anchor industry partner in September 2026.
Water is another increasingly important constraint. NUS and partners are testing water-efficient cooling and heat-rejection approaches as AI increases power density and cooling requirements. The program is examining the relationship between water usage effectiveness and power usage effectiveness, reflecting the fact that a data center can improve one metric while worsening another.
Singapore’s continued expansion of digital infrastructure gives those experiments an immediate commercial context. The country is seeking to add data center capacity while maintaining constraints around energy efficiency, sustainability and land use. Jurong Island therefore provides more than a convenient location: it can function as a controlled environment for technologies that could eventually be deployed across Southeast Asia and other hot, humid markets.
The competitive implications extend beyond Johnson Controls. Data center infrastructure providers including Schneider Electric, Eaton and Vertiv are increasingly competing around power management, liquid cooling, thermal controls and high-density AI deployments. Meanwhile, hyperscalers such as Microsoft, Google and Amazon are investing heavily in AI infrastructure, increasing pressure on the underlying facilities ecosystem to deliver more compute per unit of power.
STDCT 2.0’s value, therefore, may lie less in any single cooling technology than in its ability to test integrated infrastructure under operating conditions that laboratory demonstrations cannot fully reproduce.
For AI developers and data center operators, the central question is shifting from how much compute can be installed to how much useful compute can be delivered within available power, cooling and water constraints. Singapore’s tropical testbed offers a practical environment for answering that question—and for determining which AI infrastructure approaches are ready to move from engineering demonstrations into commercial deployments.
Market Landscape
AI is pushing data center infrastructure toward higher rack densities, greater power demand and more sophisticated thermal management. Gartner expects global data center power demand to reach 132 GW in 2026 and projects AI-optimized servers to become the largest source of server power consumption in 2027.
At the same time, operators are experimenting with direct-to-chip liquid cooling, immersion cooling, heat recovery, digital twins and grid-responsive systems. NUS’s STDCT program is explicitly designed to validate these technologies in tropical conditions, while its partnerships with major infrastructure vendors demonstrate growing industry interest in integrated AI data center design.
The market is consequently moving toward AI infrastructure efficiency as a competitive requirement alongside compute performance.
Top Insights
- AI’s rising rack densities are making cooling, power availability and water efficiency core constraints on data center expansion.
- Singapore’s tropical testbed provides a real-world environment for validating AI infrastructure beyond conventional laboratory and vendor demonstrations.
- Johnson Controls is emphasizing integrated thermal management rather than treating cooling as an isolated data center subsystem.
- Waste-heat recovery could become an important strategy for increasing usable computing capacity where grid expansion is constrained.
- Competition is broadening from AI chips and servers into the infrastructure systems needed to operate them efficiently.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
