This research develops microscopic copper wire "bridges" that improve heat transfer between computer chips and cooling systems. By reducing chip temperatures by around 3°C, the technology can lower data centre cooling energy by approximately 10%, improving efficiency and supporting more sustainable AI infrastructure.
This research improves data center energy efficiency by analyzing processor instruction sequences. By identifying and fusing recurring instruction patterns, existing general-purpose processors can execute workloads more efficiently. Even small gains at the instruction level can significantly reduce energy consumption, operating costs, and carbon emissions across large-scale data centers.