HWNEWS
The AI Grid Is Already Strained. A Solar Storm Could Make It Worse.
The North American grid watchdog has warned that rapid AI data center growth is straining the grid.
The North American grid watchdog has warned that rapid AI data center growth is straining the grid.
NERC, the North American grid watchdog, warned this year that rapid, concentrated growth in AI data-center power demand is straining the grid. That same concentration of load is now also a structural vulnerability to a severe geomagnetic disturbance — and the companies with the most to lose have barely acknowledged it.
A severe solar storm, the kind that drove currents through telegraph systems in 1859 and is generally understood to be a low-frequency, high-consequence event, can induce damaging electrical currents, called geomagnetically induced currents, through high-voltage transformers. Those transformers are among the most expensive and longest-lead-time components on the grid. With hours to days of warning, operators can rebalance load and isolate vulnerable equipment, materially reducing the risk of permanent damage. Without that warning, or with it arriving into a system already running at the edge of its tolerance, the damage can cascade and the repair cycle can stretch into years. None of that is new. The SpaceNews argument is that the math has changed: the load a major AI campus draws, comparable, the op-ed notes, to that of a small city, did not exist in its current form five years ago, and it is being added to a transmission network whose investment cycle is measured in decades.
The observational side is finally catching up. NOAA announced in May that its SOLAR-1 mission had entered a new operational era of space-weather monitoring, expanding the agency's capacity to track the solar wind and the giant plasma bursts, called coronal mass ejections, that drive geomagnetic disturbance. That is a real upgrade in the upstream signal. It is not, on its own, a fix for downstream vulnerability. A better view of an incoming storm does not shorten multi-year transformer lead times, and it does not change the basic fact that the largest new electrical loads are being added to a transmission network that was not designed for them.
This is where the accountability question gets sharp. The companies financing the AI buildout, including the largest cloud and AI infrastructure operators and the colocation providers that host third-party tenants, have every commercial reason to care about grid resilience. They also have the balance sheets to fund forecasting research, to back long-lead transformer orders, or to insist on hardened interconnections. What they have not done, at least not publicly, is articulate a position on geomagnetic risk to their own infrastructure. The companies most directly exposed to the compound scenario NERC describes have been largely silent on that specific exposure.
That silence is not necessarily negligence. The risk may be getting managed inside utility procurement teams and behind nondisclosure agreements. But it leaves a gap that is increasingly visible to outsiders. Dr. Scott McIntosh, the vice president for space operations at Lynker Space and a public-facing voice on solar-cycle prediction, has been among the researchers arguing that operational space-weather forecasting is underfunded relative to the potential damage. The argument, in plain terms, is that the cost of a bad day is large enough that the present level of preparation is mismatched to the exposure.
What changes that? Three things, in roughly increasing order of difficulty. NERC's next large-load bulletin, due in the coming year, will signal whether the AI-load volatility warning is being treated as a planning problem or a reliability problem. SOLAR-1's commissioning will determine how much additional warning time actually reaches grid operators; press-release capacity is not the same as forecast skill, and that distinction will show up in the data. The third move, and the one to watch on the industry side, is any major cloud or AI operator's energy-sourcing disclosure that names geomagnetic preparedness as a procurement criterion; that would, in a single paragraph, reframe the conversation from public-sector advocacy to private-sector capital allocation. Until one of those moves, the AI grid's space-weather exposure remains a known unknown in a system that is, on NERC's own evidence, already running hot.