AI INFRASTRUCTURE
The university, which has been around since 1887 and has seen its share of technological revolutions, is now hosting about 25,000 students.
The environmental benefits go further than just reducing carbon emissions – by displacing natural gas that would otherwise heat campus buildings, the system cuts the university’ s direct emissions while making the computing infrastructure useful rather than just necessary.
The results of AI optimising itself Perhaps the most intriguing aspect of this setup is how AI is being used to optimise AI infrastructure.
The monitoring systems use ML to predict when servers might fail as well as how different types of AI workloads will affect both computing performance and heat recovery output.
This creates a feedback loop where the infrastructure becomes genuinely intelligent rather than just automated. The system learns the thermal signatures of different research projects – training a language model( LM) produces different heat patterns than processing particle physics data – and adjusts both cooling and heat recovery accordingly. The predictive capabilities extend beyond just the
140 September 2025