How Reliable Is Fusion Energy for Powering Post-Scarcity?
Most fusion questions ask when. This one asks something more useful. A composite of the way it tends to come up:
“Fine, say fusion works and the plants get built. Can a whole society actually run on it? Solar quits at night, reactors go down for maintenance, the grid already buckles in heat waves. If food, water and housing all depend on the power staying on, what happens on the day it doesn’t?”
That is exactly the right worry, because Unscarcity asks people to treat energy like air. The book’s preamble describes a system that treats food, shelter, energy and healthcare “the way we already treat the power grid: always on, owed to everyone, with no economy attached.” A promise like that is only as good as its worst day.
Reliability here is really three questions. Will the reactor run steadily, year after year? Will the grid between the reactor and your kitchen survive what nature throws at it? And if fusion is late, or breaks, does everything else break with it? The answers differ, and the second is the one people forget.
The Reactor: Steady Fuel, Unproven Plumbing
Start with what fusion does well. A solar farm depends on the weather and a gas plant depends on a pipeline. A fusion plant needs neither. The fuel is deuterium from seawater plus lithium, and our research note on the fusion race puts the supply at billions of years. Its case for fusion is “baseload density,” power that stays on at night, and its description of the alternative is blunt: “You can’t run a hospital on ‘mostly available’ power. You can’t smelt aluminum when the sun sets.”
Where the fuel comes from matters as much as how much there is. Our Carrington analysis points out that fusion plants “don’t require the extensive logistics network that keeps natural gas plants or coal plants running.” Its summary: “When the GPS-coordinated just-in-time delivery trucks stop, your lights stay on anyway.”
The physics is no longer the open question either. The National Ignition Facility reached ignition on December 5, 2022, getting more energy out of the reaction than its lasers delivered to the target. In 2026 Helion’s Polaris became the first privately developed machine to run on deuterium-tritium fuel, and it has since reached 150 million degrees Celsius. It now runs nearly every day.
Nearly every day is a prototype’s schedule, though. A power plant has to run for decades, and that depends on hardware nobody has proven yet: the parts closest to a 150-million-degree plasma, which must survive that heat year after year while doing useful work. Chapter 1 of the book names the remaining job plainly: “making it cheap and reliable enough to deploy everywhere.”
From the Podcast: The Business End of the Reactor
The August 22, 2026 episode of Minds, Bodies & Terawatts, our daily podcast, covered one of those parts. Its headline: The First Fusion Component Test Facility in the Country Is Being Designed in a San Diego Suburb.
General Atomics is designing a facility in Poway, California, under contract with the Department of Energy, to test a blanket: a lithium-based structure lining a reactor’s inside walls. It does two jobs. It breeds the fuel that keeps fusion going, and it pulls heat out of the plasma. The engineer leading the design told NBC that nobody has ever tested a blanket component at relevant heat in a background magnetic field. He calls the blanket “the business end of the reactor.”
That detail goes straight to the reader’s question. The seawater-forever argument quietly assumes the blanket works, since it is where the reactor makes more of its own fuel. Design runs about a year, and testing should begin in three to five years, in a chamber the company already owns. The episode didn’t dress it up: “This tests a component. It isn’t a power plant.” Its reading of the whole field: “Physics risk on fusion is close to zero, because the sun settles it. What’s left is engineering risk, and that’s schedules, materials and money.”
Then came the line that matters most for this question: “Fusion raises the ceiling, it isn’t holding up the floor.”
The Weak Link Is the Wire
Suppose every reactor works perfectly. The power still has to travel, and the equipment that carries it is the most fragile part of the system.
Large power transformers weigh hundreds of tons and are custom-built. A June 2024 CISA report found the U.S. already about 30 percent short of them, and by May 2026 Wood Mackenzie was tracking lead times of roughly four years. McKinsey’s September 2026 analysis expects a 38 percent supply shortfall in North America and Europe to persist past 2030.
Then add the sun. In 1859 the Carrington Event threw sparks from telegraph equipment and set telegraph paper alight. A storm that size today would drive currents through long transmission lines strong enough to cook those same transformers. In July 2012 a Carrington-class eruption crossed Earth’s orbit, and NASA says that had it come a week earlier, Earth would have been in the line of fire. Lloyd’s of London estimates a severe storm could inflict $2.4 trillion in global losses over five years in its moderate case.
A flawless fusion plant behind a burned-out transformer delivers nothing. So the Foundation’s design doesn’t hang everything on one continental grid. It calls for a “federated mesh” of local fusion-powered microgrids that can island themselves, disconnecting from the regional network and running alone when a storm is coming. In the article’s words: “You might lose the connection to your cousin in another city for a few days. You don’t lose refrigeration for a decade.”
Consistency Comes From Redundancy
The third question, what happens if fusion is late, is already being answered in small ways. SPARC’s first plasma was a 2025 target, then a 2026 one, and now points at 2027. Helion kept its 2028 date and changed the word from “electricity” to “operations.”
The framework was built expecting that. Its energy plan has a primary path, commercial fusion, and a fallback, solar plus storage, whose costs Chapter 1 notes have fallen 90 percent since 2010. Each of the three abundance drivers, energy, robots and AI, is designed so that any one can carry 70 to 80 percent of the load if the others stumble. Fusion is what would make energy close to free and unlimited. The rest of the system does not stand or fall with it.
On unscarcity.ai: Three Articles Behind This Answer
When the Sun Ends Civilization takes the Carrington scenario apart: the transformer math, the GPS dependencies, and the four pillars the Foundation adds for resilience. Systems That Bend works through what happens if fusion stalls until 2055, driver by driver. The Fusion Revolution tracks the machines themselves, including the hedges companies have quietly written into their own schedules.
In the Book: The Human Circuit Breaker
Chapter 1 of Unscarcity, “The Foundation (Solving for Survival),” answers the reliability question with people as well as hardware. The section called “The Human Circuit Breaker” asks: “Why don’t we fear the power grid turning off and never coming back? Because a critical mass of the population knows how to turn it back on.”
It walks through Carrington in detail, then states the design assumption: “The Foundation is not built to assume none of these will happen. It is built to assume some of them will, and to keep civilization upright when they do.” The mechanism is Civic Service, a four-year commitment modeled on Switzerland’s citizen-service tradition. Automated systems handle 99 percent of routine operations; the humans cover edge cases and emergencies. Across a century, the chapter estimates, roughly 2 billion people will have served and kept the knowledge.
The section’s scene tests a different kind of reliability. In 2035 Maria’s daughter Sophia, 20, serves in the Local Energy Directorate. The grid AI flags a spike and suggests cutting power to a row of “low-priority, high-consumption” units. One of them is a community center hosting a wake for a neighborhood elder, with eighty people inside. Sophia uses her Civic Authority to tag it “Critical Infrastructure: Sanctuary Status.” The lights stay on, and the grid adjusts around the exception. The override goes on a public record, with her name on it.
More questions like this one are collected on the blog page. Follow the series on Substack.
