
What happened
ON.energy says AI data center outages are architecture failures, not supply failures, and that its medium-voltage inline power system cleared ERCOT's large-load voltage ride-through rules in an early-2026 test at the National Laboratory of the Rockies.
Why it matters
The standard power stack dates to when 'large load' meant 50 megawatts; in a 2024 Virginia event, roughly 60 facilities and 1,500 megawatts dropped at once, and the July 22, 2026 Ashburn fault knocked over 3 gigawatts off the grid.
What to watch
ON.energy's claim hinges on whether operators certify one medium-voltage box instead of untangling every transformer and UPS, and whether such equipment earns tax credits and grid-program revenue. Watch the July 22, 2026 Ashburn case.
WHO IT HITSOperators of AI data center campuses and the utilities that interconnect them face the architecture mismatch the article describes, since each new campus must clear large-load ride-through rules; ON.energy's pitch is that certifying one medium-voltage box could shorten permitting timelines.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
The article frames the AI power debate as mostly about generation—more turbines, more solar, more transmission—and then argues the Virginia outages were not supply failures but architecture failures. The grid, it says, was built around predictable loads like steel mills, refineries, and houses at dinnertime, while AI campuses can swing 70% of their load in milliseconds and trip offline at the first sign of trouble upstream. Two events anchor the argument: a 2024 Virginia event where a single failed surge arrester dropped roughly 60 facilities and 1,500 megawatts, and the July 22, 2026 Ashburn fault that knocked more than 3 gigawatts of load off the grid in seconds.
The proposed fix has three parts—move protection up to medium voltage, move it out to modular enclosures near the substation, and move it into the path so nothing filters around it. The article says this rewrites downstream line items: the utility certifies one medium-voltage box instead of untangling every transformer, UPS, chiller, pump, and switchgear lineup, and engineers can swap chip generations without a fresh interconnection study. Inside the fence, UPS rooms become compute or cooling space, and the economics shift because equipment that runs at medium voltage, sits outside, and stores its own energy can qualify for tax credits and earn revenue in grid programs like peak shaving and demand response.
The stakes hinge on whether this architecture actually clears utility certification in practice and whether the tax-credit and grid-program economics hold, since those are the two levers the article says turn backup power from insurance into something that pays for itself. For operators planning the next wave of AI campuses at gigawatt scale, the question is whether they arrive as a strain on the grid or as strength for it—and the article notes the industry has not yet named this layer, which ON.energy calls the medium-voltage AI UPS.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Michael Burry, famous for The Big Short, is short Nvidia, Palantir and Tesla, and in his Substack newsletter s…

Investors have three creative routes to Anthropic exposure before its expected IPO: buying Alphabet, Amazon, o…

Much of the attention on AI infrastructure buildouts is now tied to sheer compute power, with dominance define…

Barron's reported September 10 that Kepler Computing emerged from stealth with a memory architecture using fer…

Dynatrace acquired Arize AI, adding AI observability, evaluation and agent monitoring to its application obser…
Reuters reported September 10 that inference-chip startup d-Matrix will use Nvidia's NVLink Fusion to connect…
