MIT taps Ferveret cooling platform for AI research on nuclear materials

MIT's Materials in Extreme Environments Lab is using Ferveret's Adaptive Phase Cooling platform to train AI models and machine-learning interatomic potentials for nuclear materials research. Ferveret claims up to 15% more server-level compute and up to 35% with facility-level gains, though no independent benchmark of these figures has been published.

• MIT uses Ferveret's platform for AI analysis of coherent X-ray data on nuclear materials
• Ferveret claims up to 15% more server compute and up to 35% with facility gains
• Cooling relies on subcooled boiling with local bubble condensation
• The 15% and 35% figures are company calculations without a published benchmark protocol