Microsoft model flags grid risk from space weather across 66,935 substations

The system covers 66,935 substations in the continental United States. Microsoft says it detected nearly 80 percent of major space weather events during its evaluation period. The company published the details on 30 September in a research blog describing work done during a summer internship.

Solar wind observations feed the pipeline, which also uses forecasts of two geomagnetic indices, the Auroral Electrojet and Disturbance Storm Time. Local latitude, bedrock geology and ground conductivity data are then added to produce location specific risk estimates. The post explains that resistive bedrock can produce stronger geomagnetically induced currents than conductive geology. That is why the same storm can hit two substations differently. It matters because induced currents in transmission networks can damage transformers and raise operational risk.

The problem is not only knowing a storm is coming. It is estimating when and where the effects will be worst while there is still time to act.

Microsoft's framing is consistent with how utilities behaved during the May 2024 geomagnetic storm, when auroras appeared far beyond their usual range and operators across North America prepared for possible impacts. That storm degraded GPS accuracy and satellite operations, with knock on effects for farm equipment in North America and Europe, according to the research post.

Space weather research is a crowded field, and this is one contribution among many. To be clear about what the evaluation period covered and how the 80 percent figure was measured, readers have to go to the blog itself, which does not give either in the summary provided. The company also does not say when the system might move from research to operational use, or which grid operators, if any, are testing it.

Other recent work in the same dossier shows how quickly the commercial side of space is moving. On 30 September, Tech.eu reported that Lithuania's Blackswan Space closed a 2.5 million euro round led by Iron Wolf Capital, bringing total funding to 3.2 million euros, to fly autonomous navigation hardware in 2027 and support Europe's first in space refuelling mission a year later. SpaceNews reported the same day that AstroForge plans to fly a spacecraft next year testing full autonomy using artificial intelligence.

Russia's long delayed crewed spacecraft also resurfaced this week. Ars Technica reported on 30 September that Roscosmos chief executive Dmitry Bakanov has dropped the Eagle name, referring to the vehicle as PTK NP, and that parachute drop tests are due this autumn in the Far East. The same report says Russian officials are holding to a debut crewed launch in 2028 or 2029, likely on an Angara A5M rocket, and notes that similar dates have slipped before.

None of that changes the basic grid problem. A storm does not need a new spacecraft to cause damage, and the warning window Microsoft describes is short: half an hour to an hour. Whether that is enough for operators to react is a question the research post does not answer.