7 min readScienceResearch

Space weather and our infrastructure

A giant storm from the Sun could cause real trouble for power, satellites, GPS and radio, but the very worst scenarios are far less certain than headlines suggest.

ScienceResearch

A giant storm from the Sun could cause real trouble for power, satellites, GPS and radio, but the very worst scenarios are far less certain than headlines suggest.

The big picture#

"Space weather" means eruptions on the Sun that reach Earth. Our modern world depends on things these eruptions can disturb: electric grids, satellites, navigation and long-distance radio.

Most storms do little harm. The ones worth planning for are rare, and this guide explains what is known, what is guessed, and what is only rumor.

How a storm travels from the Sun to Earth#

Two kinds of eruption matter, and they arrive at very different speeds.

  • Solar flare: a burst of light and X-rays. It reaches Earth in about 8 minutes and can block some radio signals on the sunlit side of the planet.
  • Coronal mass ejection (CME): a huge cloud of electrically charged gas thrown off the Sun. The fastest ones reach Earth in 15 to 18 hours. Slower ones take several days.

Whether a CME does damage depends mostly on its magnetic field. If that field points south, against Earth's own north-pointing field, the two connect and energy pours into our surroundings. If it points north, much less happens.

The US weather service (NOAA) rates geomagnetic storms from G1 (minor) to G5 (extreme). Stronger storms are much rarer. The counts below are per 11-year solar cycle.

  • G1 (Minor): 1,700 per cycle.
  • G3 (Strong): 200 per cycle.
  • G5 (Extreme): about 4 per cycle.

What a storm can break#

The power grid is the biggest worry, because almost everything else depends on electricity.

  • Power grid: Overloaded equipment switches off parts of the grid. Real example: Quebec, March 1989.
  • Satellites: Thicker upper air drags them down; electronics get damaged. Real example: Starlink loss, February 2022.
  • GPS: Signals scramble; precision users lose their position. Real example: Farm equipment, May 2024.
  • Flights and radio: Polar routes lose radio contact and may be rerouted. Real example: Polar flights, May 2024.

How the grid gets hurt. A storm makes the magnetic field near the ground change quickly. That creates weak electric currents in the Earth, and long power lines collect them like antennas. The currents flow into big transformers, which overheat and waste power. Safety equipment may then switch parts of the grid off.

Quebec, 1989. The grid collapsed in about 90 seconds. Six million people lost power for about nine hours. The main cause was equipment tripping offline, not transformers being destroyed, although one large transformer at a New Jersey power plant did fail.

Why transformers matter. They are hard to replace. New large units took about 120 weeks on average in 2024. Estimates of how many a giant storm would damage disagree widely:

  • One insurance-market study (Lloyd's) put it at about 10 to 20 for a storm stronger than Quebec's.
  • One US study counted more than 300 units "at risk" in its highest-end scenario.
  • The grid regulator NERC judged that a large number failing is unlikely.

Satellites. In February 2022 a modest storm made the upper air thicker, and 38 of 49 newly launched Starlink satellites were lost. In May 2024 nearly 5,000 satellites adjusted their orbits, but Starlink reported no losses.

GPS. In May 2024 high-precision farm GPS was unusable for up to 37 hours in northern Norway. Position errors reached about 60 meters in parts of the Americas. Some US farmers paused planting. One estimate puts the cost at about $500 million, but it rests on a single source.

Scary claims with no proof. You may have read that undersea internet cables would be destroyed, financial markets would halt, or all flights would stop. No past storm has done any of these. For cables, one study assumed some risk, while Google researchers concluded long-distance cables would survive even a Carrington-sized storm.

Storms from the past#

The strongest storm ever measured directly hit in 1859, and the 2024 storm was less than half as strong.

Scientists compare storms using a magnetic measure called Dst, in nanotesla (nT). It shows how much a storm weakens Earth's magnetic field. The more negative the number, the stronger the storm.

  • Carrington, 1859: strength (Dst, nT) about -900 (estimates -750 to -1,100). Telegraph shocks and fires; auroras seen near the equator.
  • May 1921: strength (Dst, nT) about -907. Known as the New York Railroad storm.
  • March 1989: strength (Dst, nT) -589. Quebec blackout.
  • Oct-Nov 2003: strength (Dst, nT) about -383. A GPS-based aircraft navigation service was out for about 11 hours.
  • May 2024: strength (Dst, nT) about -406. Satellite maneuvers, farm GPS problems, no major blackouts.

In July 2012 a huge eruption was aimed away from Earth and missed us. Modeling suggests it could have reached -500 or stronger. The Carrington storm took about 17.5 hours to arrive.

How often does a Carrington-sized storm happen? An insurance-market study (Lloyd's, 2013) estimated about once every 150 years, with a range of 100 to 250 years, based on historical aurora records. For a Quebec-sized storm it estimated about once every 50 years. These are rough estimates, not forecasts.

Could a much bigger storm happen?#

Tree rings show that particle bursts from the Sun, probably larger than anything in modern records, happen about once every 1,500 years, at irregular intervals.

In 2012 Japanese physicist Fusa Miyake found a sudden jump in radioactive carbon (carbon-14) in tree rings from the years 774-775 CE. Cosmic particles hitting the upper air make carbon-14, and trees take it in as they grow. A jump in one year's ring is a record of a sudden burst of particles.

Since then, researchers have found about seven such "Miyake events" in the past 15,000 years, plus two possible ones. The strongest known, from about 12,350 BCE, was found in pine trees from the French Alps. Its carbon-14 jump was about 38 parts per thousand, compared with about 12 for the 774 CE event.

Two cautions apply:

  • These are bursts of particles. Scientists do not know how to translate them into a Carrington-style storm, and the 1859 storm left no clear mark in tree rings or ice.
  • The cause is still debated.

"About once every 1,500 years" is an average, not a schedule. It does not mean one is due.

What can be done, and how worried to be#

Good warnings and careful grid operation can prevent most of the damage, but only if the preparation happens before the storm.

  • Warnings. Satellites about a million miles toward the Sun measure the solar wind and give 15 to 60 minutes of notice. NOAA's newer satellite SOLAR-1, launched in September 2025, is now its main source. A CME itself can be seen leaving the Sun, which gives hours to days of notice.
  • Grid operators. Before a storm they can bring extra power reserves online and reduce load on exposed lines. North American rules (NERC standard TPL-007) require utilities to test their grids against a benchmark storm.
  • Hardware. Devices that block the ground currents help. They must be placed carefully, because blocking at one site can push currents onto neighbors.
  • Satellites. Operators can put them in a protective "safe mode" ahead of a storm.
  • Policy. The US issued an executive order on space weather in 2016 and passed the PROSWIFT Act in 2020. The UK has listed severe space weather in its national risk register since 2011.

Estimated costs vary enormously. Lloyd's estimated $0.6 to $2.6 trillion for a Carrington-sized storm hitting the US. Another study (Oughton et al., 2017) estimated $6.2 to $42 billion of lost US economic output for each day of blackout, in 2011 dollars.

So how worried should you be? Concern is reasonable, but panic is not. A regional blackout like Quebec's is the best-documented risk, and utilities plan for it. A years-long worldwide collapse is a worst-case scenario, not a forecast, and the evidence behind it is thin.

How sure we are, and sources#

This is a plain-language version of a fact-checked revised edition, and a few figures are still being confirmed.

  • The storm-frequency counts match NOAA's own published table (version dated December 11, 2023), checked on October 30, 2026.
  • The 1989 and 2003 storm strengths match Kyoto's final data. The May 2024 value is Kyoto's provisional figure, and other analyses give up to about -412 nT.
  • Transformer damage estimates differ a lot between studies, as shown above.
  • The sixfold rise in air density at satellite height in May 2024 is a model result, not a measurement. It is left out here for that reason.
  • The $500 million farm cost has one source.

Main sources (the technical edition has the full list of 70):

Esc