Synopsys has joined a Singapore research consortium led by the National University of Singapore (NUS) to investigate how artificial intelligence data centers can be cooled more efficiently in tropical conditions, the company announced on October 28. The work will combine engineering simulation with physical tests rather than rely on computer modeling alone.
The collaboration is part of the Sustainable Tropical Data Centre Testbed Phase 2.0 (STDCT 2.0), hosted by the university’s College of Design and Engineering. Synopsys plans to contribute tools for modeling heat transfer and fluid flow, as well as technical expertise, training and research mentorship.
Testing liquid cooling for high-density AI systems
Higher-density racks built for AI accelerators produce substantial heat in a concentrated space. In Singapore, high ambient temperatures and humidity can make conventional cooling systems less efficient. Synopsys said cooling can account for as much as 40 percent of energy consumption in a tropical data center, an industry estimate that does not represent measured savings from this particular collaboration.
Among the initial research topics is two-phase direct-to-chip cooling, in which a fluid absorbs heat near high-output computing components and changes phase during the process. Researchers intend to investigate how a design behaves under different loads and operating conditions before comparing model predictions with measurements from the testbed.
Other planned activities include AI-assisted design studies and reduced-order models, which simplify detailed physics simulations so they can run faster in operational settings. The partners want to use those models in digital twins that can estimate temperatures and support decisions about liquid-cooled infrastructure.
Digital twins are useful only to the extent that their predictions match real conditions. The proposed program therefore includes physical validation and techniques for representing uncertainty in the modeled results. No deployment date, commercial product launch or verified energy-saving result was disclosed in the announcement.
Part of a larger Singapore research program
NUS and Singapore industrial-land agency JTC had previously outlined plans for a multi-megawatt, AI-ready research facility on Jurong Island, where sustainable power and cooling technologies could be evaluated before wider commercial adoption. The testbed’s research scope includes heat rejection, energy efficiency, water use and integration with lower-carbon infrastructure.
Synopsys is not the only commercial participant. TNGlobal reported in September that Johnson Controls had joined STDCT 2.0 to explore thermal management and smart controls. Its modeled energy-efficiency estimates were distinct from measured outcomes at the university’s testbed.
Johnson Controls joins NUS tropical data center testbed for AI infrastructure