The Sun could knock out the grid the AI boom is built on — and almost no one is planning for it

Space 5 min 80 sources

A solar storm severe enough to damage the grid is overdue, and the data centers powering AI can't ride one out. Plus the loudest black-hole crash ever heard, Euclid's stunning Milky Way core, and a repair spacewalk that says a lot about a tired space station.

Key takeaways

A severe solar storm could wreck the hard-to-replace transformers the power grid runs on, and the AI data centers now straining that grid are the least able to survive a long blackout.
Gravitational-wave detectors caught the loudest black-hole collision yet, and scientists pulled from it the first direct read of a black hole's event horizon — matching Einstein exactly.
Europe's Euclid telescope imaged 60 million stars in the Milky Way's core, opening a new stretch of planet-hunting, while a routine ISS repair spacewalk underlined how old the station is getting.

The biggest space story this week wasn’t a launch. It was a warning about the Sun, and what it could do to the ground.

A storm we can see coming, and barely prepare for

The Sun throws tantrums. Every so often it hurls a billion tons of charged particles toward Earth in a burst called a coronal mass ejection — a CME, a cloud of solar gas flung out by an eruption on the Sun’s surface [2]. When that cloud slams into Earth’s magnetic field, it can push electrical currents into power lines that the grid was never built to carry. Those stray currents overheat and wreck high-voltage transformers — the room-sized machines that step electricity up and down across the network, and among the most expensive, hardest-to-replace parts of it [2].

We know this can happen because it already did. In 1859 a storm now called the Carrington Event — one of the strongest on record — knocked out telegraph lines around the world [2]. A storm that size today would hit a planet wired to depend on electricity for almost everything. Estimates of the economic damage from an extreme solar storm run into the trillions of dollars, and replacing a wave of fried transformers can take months or years, because you can’t keep many spares of a custom-built machine on a shelf [2].

This week the warning got sharper because of what we’ve started plugging in. AI data centers are spreading fast, some drawing as much power as a small city, and they are straining the grid even on a calm day [2]. The North American Electric Reliability Corporation — the body that watches grid stability — recently flagged that the huge, sudden power swings from training AI models are a “high likelihood, high impact” threat on their own; during a training run, demand can jump by hundreds of megawatts in an instant [2]. Stack a solar storm on top of an already-stressed grid and the risks multiply.

And data centers are the worst-placed to ride it out. Most of us can lose power for a few hours. A data center can’t: even a brief interruption can corrupt the work in progress, and no bank of backup batteries lasts through a blackout measured in weeks [2].

There is some movement. In 2025 NOAA — the U.S. weather agency — declared its first satellite built only for watching space weather, SOLAR-1, operational; parked between Earth and the Sun, it gives faster warning of incoming storms [2]. But it only sees the Earth-facing side of the Sun, so an active region can rotate into view with little notice [2]. Space-weather forecasting is nowhere near as sharp or as far-ahead as the forecast that tells you it’ll rain tomorrow [2]. The author of this week’s argument — a solar physicist — says the tech giants building the data centers should be the loudest voices pushing to fix that, and mostly aren’t [2].

The loudest crash ever heard from a black hole

Far from the Sun, a different kind of detector did something remarkable. Back in January 2025, the gravitational-wave observatories LIGO, Virgo and KAGRA — instruments that feel space itself stretch and squeeze as ripples pass through — caught the strongest signal they’ve ever recorded, from two black holes about 32 times the Sun’s mass crashing together [58]. Gravitational waves are tiny tremors in space, set off when very heavy objects accelerate hard; this collision rang space like a struck bell [58].

This week a team published what they pulled out of that signal: a faint feature, the “direct wave,” that carries an imprint of the merged black hole’s event horizon — the surface of no return, the boundary past which not even light gets back out [32][58]. Measurements matched what Einstein’s relativity predicts for a spinning black hole exactly [32]. Researcher Neil Lu put it plainly: “We measured the last sound the black holes made when they crashed” [58]. The point isn’t one event. It’s that gravitational waves now look like a tool for studying the edge of a black hole directly — a place we will never visit and never photograph up close [58].

60 million stars, and a hunt for new worlds

Europe’s Euclid telescope released the largest, most detailed image ever taken of the visible light pouring out of the Milky Way’s crowded core, picking apart more than 60 million individual stars in the dense region called the galactic bulge [33][53]. Euclid was built to map dark matter and dark energy — the invisible stuff thought to make up most of the universe — not to count stars [33]. But it turned out to be unusually good at it [33].

The practical payoff is planet-hunting. Astronomer Eamonn Kerins of the University of Manchester said the data “fires the starting pistol” on a new stretch of exoplanet discovery — possibly pushing the count of known worlds beyond our solar system from about 6,000 today toward more than 100,000 [33]. NASA’s Roman telescope will survey the same crowded core later, and Euclid’s image is the preview of what that will see [21].

A tired arm, on a tired station

The quietest story said the most. NASA astronauts Jessica Meir and Chris Williams are set to spacewalk on June 30 to replace a faulty wrist joint in Canadarm2, the station’s main robotic arm, after engineers noticed it drawing too much current and not moving as expected [11]. The arm was designed from the start with swappable parts, and a spare joint was already aboard — so this is maintenance working as planned, not a crisis [11].

But it landed alongside warnings from NASA’s own safety advisers about the long-term health of the International Space Station and its aging spacesuits [11]. The station has been continuously crewed for a quarter of a century. A swapped joint is routine; the worry underneath is that everything up there keeps getting older, and the bill for keeping a 25-year-old machine alive in orbit only grows [11].

02 · Lesson · why it matters

Why we keep building taller towers on the same shaky ground

The risks that wreck us are rarely the loud ones. They are the quiet, rare ones we keep treating as someone else's problem — while stacking more and more weight on top of them.

A danger that almost never arrives

The Sun has been throwing storms at Earth for as long as there has been an Earth. Most miss us or pass harmlessly. Once in a long while, one is big enough to push currents into the power grid and burn out the transformers that hold it together. The last one that size hit in 1859, before anything ran on electricity. It blew out telegraph wires and then went away.

So the threat has a strange shape. It is real, it is physical, and it is also almost never here. On any given day — any given decade — the safe bet is that nothing happens. And a bet that pays off every single day is a very hard bet to argue against. Why spend money guarding against a storm that hasn’t come in your lifetime, your boss’s lifetime, or your grandparents’?

That is the trap. The danger is rare enough to ignore and severe enough to be ruinous. We are very good at the first half and very bad at the second.

The cost of a quiet risk is paid by no one — until it’s paid by everyone

Here is the thing about a risk like this: nobody owns it.

The power company plans for the storms it sees every year, not the one every century. The data-center operator plans for the outages that happen on Tuesdays, not the one that knocks out a region for months. The chip-maker, the AI lab, the city that depends on the grid — each looks at the rare disaster and reasonably concludes it is somebody else’s job. The forecasters who could warn us are funded as if the warning barely matters, because most years it doesn’t.

So the cost of preparing sits in a gap between everyone, and falls through. This is what a quiet risk does: it spreads the blame so thin that no single hand has to hold it. Each player is being sensible. The system as a whole is being reckless. Both are true at once, and that is exactly why it doesn’t get fixed.

We keep raising the stakes on the same ground

Now add the part that makes this week’s news different from 1859.

We are not building on the grid the way we used to. We are building harder. AI data centers draw the power of small cities and run without pause; a brief flicker can corrupt a week of work. They are, by design, the things least able to survive a blackout — and we are bolting more and more of them onto a grid that was already straining before they arrived.

Watch the move, because it repeats everywhere. We find something useful — cheap electricity, fast networks, a single shared signal. We come to depend on it. Then, because it works, we build the next thing on top of it, and the next, each one assuming the floor will hold. The tower gets taller. The foundation does not get stronger. And the fall, when it comes, is no longer the height of 1859. It is the height of everything we have stacked since.

A rare risk doesn’t stay the same size while we ignore it. It grows — not because the Sun changes, but because we keep increasing what’s resting on the ground beneath us.

Who wasn’t in the room

There’s a quieter shape under all this. Whether we prepare for the storm was mostly decided years ago, by people choosing budgets and priorities — and the choices look like plain facts now. “Space-weather forecasting gets a small fraction of the money weather forecasting gets” sounds like a law of nature. It is a decision. Someone, at some point, weighed the rare disaster against the certain cost and quietly chose to underspend.

That choice served the people making it — every year the storm didn’t come, they were right, and the money went to louder problems. It also left everyone downstream exposed, including people who never knew the choice was being made. The reader running a phone, a bank account, a job that lives on a server is standing on a floor whose strength was decided by someone they’ll never meet, optimising for a risk they couldn’t feel.

That isn’t a villain. It’s how rare risks get handled by default. Name it, though, and the “natural fact” turns back into what it always was: a bet, placed on our behalf, that we mostly don’t know we’re holding.

What this leaves you with

It’s tempting to walk away from this feeling clever — to look at the grid, the data centers, the underfunded forecasts, and see the whole creaking machine clearly.

But the honest end is the harder one. You are not standing outside this tower, watching it sway. You are living on a middle floor. Almost everything you rely on rests on systems you didn’t design, can’t see, and couldn’t fix — and so does everything the people above and below you rely on. The grid operator can’t see the AI lab’s exposure. The AI lab can’t see the forecaster’s blind spot on the far side of the Sun. No one is looking at the whole tower, because no one can.

That’s the part to carry. Not “those people should prepare better” — though they should — but the smaller, steadier truth underneath it: we are all standing on ground we trust without being able to check, building higher on a bet we didn’t place. Knowing that doesn’t make the floor more solid. It just makes you hold your certainty about it a little more loosely.

03 · Lab · your turn

The Storm That Hasn't Come

Rehearse betting against a rare, ruinous risk year after year while the stakes quietly grow, and feel why doing nothing looks smart until the year it isn't.

04 · Hope · carry this

A storm we can name is a storm we can prepare for, and the fact that scientists are watching the Sun, building the early-warning satellites, and saying out loud what the rest of us missed is exactly how a quiet danger becomes a problem we solve in time.