Summary

• The Met Office has completed the second phase of its Microsoft-managed cloud supercomputing programme and migrated its Managed Archive Storage System from Exeter.
• Its expanded platform uses approximately 1.8 million processor cores, supports around 60 quadrillion calculations per second, and processes more than 50 billion weather observations daily.
• The organisation projects £11.3bn in wider socioeconomic benefits from improved forecasting and climate services, although those benefits have not yet been fully realised or independently quantified.

The Met Office has completed the second phase of its cloud supercomputing programme, expanding the processing capacity available for weather and climate modelling while transferring its scientific data archive from Exeter into a managed cloud environment.

The upgrade builds on the transition of the UK’s operational forecasting systems to a Microsoft-managed platform in May 2025. That initial deployment maintained the capabilities of the previous supercomputer, while the newly completed phase introduces substantially greater computing capacity for scientific development and more demanding simulations.

The expanded system uses approximately 1.8 million processor cores and can perform around 60 quadrillion calculations per second, according to the Met Office. It processes more than 50 billion weather observations daily, supplying the models behind forecasts, severe weather warnings, and climate information used by public services and businesses.

The organisation has also completed the migration of its Managed Archive Storage System, known as MASS. The move brings an important store of historical and operational scientific information into the cloud environment, creating additional capacity for the rapidly growing volumes of data associated with numerical forecasting and climate research.

The archive moves alongside forecasting systems

Weather forecasting depends on a continuous flow of information from observing stations, satellites, radar systems, aircraft, and other sources. Numerical models combine those measurements with mathematical descriptions of atmospheric behaviour to calculate how weather conditions are likely to develop.

The resulting forecasts generate substantial quantities of data, while previous observations and model outputs remain valuable long after an individual forecast has been issued. Scientists use historical information to evaluate new forecasting methods, investigate unusual events, and measure changes in the performance of numerical models.

The Met Office’s Managed Archive Storage System supports those activities by retaining information required for forecasting operations and research. Its transition from the previous Exeter-based archive forms a significant part of the latest infrastructure milestone.

As models become more detailed, their storage requirements increase. Dividing the atmosphere into smaller computational regions produces larger datasets, while running more simulations creates additional information that researchers may need to retain for later analysis.

Moving the archive into the managed cloud environment is intended to accommodate that growth without relying on the same local storage arrangements as the previous system.

However, access to large volumes of data remains an engineering consideration. Storage capacity alone does not determine how quickly scientists can retrieve information, and applications performing intensive calculations may also depend on network throughput, data organisation, and efficient movement between storage and processing systems.

The Met Office has not disclosed a detailed comparison of archive retrieval times or storage operating costs following the migration. Those measures will influence how effectively researchers can use the additional resources, particularly where models require repeated access to extensive historical datasets.

More computing supports finer scientific models

The additional processing capacity allows scientists to investigate improvements that would have been difficult or expensive within the previous operating environment. These include higher-resolution simulations, more complex representations of atmospheric processes, and larger collections of model runs used to examine uncertainty.

Numerical weather prediction divides the atmosphere into a computational grid and estimates how conditions evolve over successive time steps. Smaller grid cells can represent local variations in greater detail, although they also increase the number of calculations required.

Higher resolution can help improve the representation of rainfall, cloud formation, and other atmospheric processes that vary over relatively short distances. The Met Office says the upgraded capability is already supporting improvements to how its models represent rainfall and incorporate observational data.

Such improvements can be valuable when forecasting weather that affects relatively small geographical areas. Heavy rain or localised wind conditions may have substantial consequences for transport and infrastructure even when the wider regional forecast appears less severe.

Greater computing power also supports larger ensembles, in which scientists run several simulations with differing initial conditions or model configurations. Comparing their results helps assess uncertainty and the range of plausible outcomes, rather than relying entirely on a single forecast.

These methods are important because even small uncertainties in atmospheric observations can influence the development of weather systems. An ensemble provides additional information about the confidence that can reasonably be placed in a prediction.

For climate research, larger and more detailed ensembles can support investigation of longer-term changes and possible extremes. The additional resources can also help researchers examine how different model assumptions affect results over extended periods.

Computing capacity does not independently guarantee greater forecast accuracy, however. The quality of the observations, scientific methods, and physical processes represented within a model remains important, while changes must be evaluated against real weather and established benchmarks.

Managed infrastructure changes operational responsibilities

The supercomputing platform is operated through a partnership with Microsoft, combining the company’s cloud infrastructure and operational services with the Met Office’s scientific expertise.

Under the managed arrangement, Microsoft provides round-the-clock operational capability and engineering support, while Met Office specialists concentrate on the forecasting systems, scientific models, and services delivered using the infrastructure.

The Met Office says reliability and resilience have improved under the new arrangements, although its latest update does not publish a detailed comparison of availability or recovery performance between the previous and current platforms.

For a national forecasting organisation, reliability involves more than keeping computing equipment available. Forecasts must be produced within operational deadlines, requiring observations, storage, numerical models, networking, and downstream information services to function together.

Delays in one part of that process can affect the availability of information even when the central supercomputer remains operational. Managed infrastructure therefore needs to support the wider scientific workflow rather than treating computing capacity as an isolated service.

The phased transition was designed to protect continuity. Moving the operational forecasting service first allowed the organisation to establish the new computing environment before expanding its capabilities and replacing the archive.

That sequencing reduced the need to introduce every major infrastructure change simultaneously, although the Met Office has not published a complete account of the engineering costs or challenges encountered during the migration.

Supplier dependence remains a consideration

Using a managed cloud platform provides access to specialised infrastructure and operational expertise without requiring the Met Office to maintain every aspect of the computing environment independently.

It also establishes a long-term relationship with an external technology provider. Scientific applications, archive arrangements, operational procedures, and specialised computing requirements can become closely connected to the infrastructure supporting them.

Changing suppliers in future could therefore require substantial engineering, even where the organisation retains control over its scientific information. Moving large datasets is only part of the work involved in transferring a complete forecasting environment.

Those dependencies need to be assessed alongside the advantages of the current arrangement, including access to computing capacity, operational support, reliability, and the ability to introduce further upgrades.

The Met Office describes its system as the world’s first fully managed cloud supercomputing service for weather and climate science. That characterisation comes from the organisation itself; the latest operational milestone does not independently establish how its service compares with every other national meteorological computing arrangement.

Regardless of that distinction, the completed migration demonstrates that the Met Office is now operating both its core forecasting infrastructure and scientific archive through the managed cloud model.

Projected benefits extend beyond forecasting

The Met Office expects the supercomputing programme to deliver £11.3 billion in socioeconomic benefits through improvements to weather and climate information.

That figure represents a projection of wider benefits rather than an audited total of realised savings. The latest update does not provide a detailed breakdown of the economic assumptions or the amount of benefit already delivered.

Improved forecasting can support decisions in aviation, where operators need to plan flights and manage disruption, and in electricity systems, where weather affects renewable generation and demand.

Transport authorities, emergency responders, and local government also rely on timely information when preparing for flooding, storms, high temperatures, and other hazards.

For these organisations, the value of a forecast depends not only on its accuracy but also on whether they can act on the information. An earlier warning may create opportunities to protect equipment or adjust operations, although the benefits depend on the decisions taken and the resources available.

Climate services support a different set of decisions, including long-term infrastructure planning, flood risk assessment, and adaptation to changing environmental conditions. More detailed projections could help organisations examine risks over the useful life of major investments.

Quantifying those benefits involves assumptions about the quality of future predictions and the actions users will take in response. The £11.3 billion estimate should therefore be treated as a programme forecast rather than a guaranteed financial return.

Efficiency and future scientific development

The Met Office also identifies energy efficiency as an advantage of its cloud supercomputing arrangements, although the latest update does not provide independently verified comparative figures for electricity consumption or lifecycle emissions.

Computing efficiency must be considered alongside the amount of scientific work performed. A more efficient platform may consume less energy for an equivalent calculation while total demand increases if scientists use the additional capacity for larger or more frequent simulations.

The expanded infrastructure also provides opportunities to investigate machine learning techniques alongside established numerical weather prediction. AI approaches can support selected scientific tasks, but their results must be evaluated against observations and established modelling methods, particularly during unusual or extreme conditions.

The archive migration and second-phase expansion complete the latest operational stage rather than the development of the Met Office’s scientific capabilities. Researchers can now work with substantially greater computing capacity and a cloud-based archive as they develop and test future forecasting systems.

The resulting gains will become clearer through changes in forecast performance, operational reliability, and the usefulness of the information delivered to organisations managing weather and climate risks.