AI news

1. Google DeepMind and ITER Achieve Fusion Milestone via "Kinetic AI"

Date: July 17, 2026

Google DeepMind and the ITER Organization have announced a breakthrough in nuclear fusion stability using a novel "Kinetic AI" control framework. The system employs deep reinforcement learning to predict and neutralize plasma instabilities in real--time, enabling a record--breaking 30--minute sustained high--pressure discharge. This transition from experimental physics to AI--driven engineering is expected to accelerate the commercialization of fusion power by a decade. Experts note that the system's ability to manage magnetic fields at microsecond scales effectively solves the "plasma disruption" bottleneck that has hindered fusion for decades.

Metric Specification
Sustained Duration 30 Minutes
Control Latency < 50 Microseconds
System Architecture Deep Reinforcement Learning

Source: Nature Portfolio