On 2 March 2026 the Deutscher Wetterdienst put AICON, a global machine-learning weather model running at 13 km resolution, into full operational use. Ten months earlier, in May 2025, the Met Office had moved its operational production onto a cloud-based supercomputing service supplied by Microsoft under a ten-year managed-service contract valued at £1.034 billion, and France estimates the cost of renewing Météo-France’s computing capacity over 2025–2033 at €415 million. These are the visible signs of a technological transition in European meteorology, and they invite a simple reading: better models and more computing mean better military weather support. For an armed force, however, none of these assets is independently equivalent to a forecasting capability. What matters operationally is a permanent service that ingests observations, constructs an analysed state of the atmosphere and ocean, runs and verifies forecast systems, translates their output into environmental intelligence, distributes it through authorised networks, keeps qualified people in place and continues to operate when models, infrastructure or suppliers change. An AI forecast that performs well in a scientific benchmark may still lack variables, calibration, local skill, interfaces or an accepted place in a military decision process. The unresolved question is where, in each national chain, scientific capability becomes an authoritative military service, and where that conversion can still fail.