Fusion power plants aren’t an option yet because the engineering required to turn a proven physics reaction into a reliable electricity source remains unsolved on multiple fronts. Scientists have demonstrated that fusion works: in late 2022, the National Ignition Facility produced more energy from a fusion reaction than the lasers delivered to the fuel. But generating a burst of energy in a lab and running a power plant 24/7 for decades are fundamentally different problems. The gap between them involves fuel supply, materials that can survive the environment inside a reactor, heat management, cost, and the sheer difficulty of sustaining the reaction continuously.
The Energy Gain Problem
Fusion’s central metric is called the Q factor: the ratio of energy a reaction produces to the energy put in. A Q of 1, called scientific breakeven, means the plasma produced as much energy as it absorbed. That milestone has been crossed. But scientific breakeven is misleading because it only counts the energy delivered directly to the fuel, not the vastly larger amount of energy needed to power the lasers, magnets, cooling systems, and everything else in the building.
What actually matters for a power plant is “wall-plug gain,” which compares the total electricity produced to the total electricity consumed by the entire facility. For a fusion plant to be useful, it needs to produce several times more energy than it uses. ITER, the massive international tokamak under construction in southern France, was originally designed to achieve a plasma Q of 10, meaning ten times more fusion energy out than heating energy in. But even that wouldn’t translate to wall-plug gain above 1 without major efficiency improvements across the whole system. SPARC, a smaller tokamak being built by Commonwealth Fusion Systems, is targeting a plasma Q of at least 2. Neither machine is designed to generate electricity; they exist to prove the physics works at scale.
No Natural Supply of Fuel
Fusion reactors are designed to fuse two hydrogen isotopes: deuterium and tritium. Deuterium is abundant in seawater, but tritium is radioactive with a half-life of 12.3 years, which means it decays too quickly to accumulate naturally. There is essentially no natural supply. Today’s tritium comes as a byproduct from certain types of fission reactors, and the global inventory is small.
A commercial fusion plant would need to manufacture its own tritium on-site using a component called a tritium breeding blanket. The blanket surrounds the reactor and contains lithium, which produces tritium when struck by neutrons escaping the fusion reaction. The blanket must breed at least as much tritium as the reactor consumes, and designs typically aim for about 10% extra to cover radioactive decay, processing losses, and startup fuel for future plants.
This sounds straightforward in principle, but every candidate blanket material comes with serious engineering headaches. Liquid lithium is chemically reactive and burns on contact with air or water. Lead-lithium alloys are extremely heavy and corrode structural materials. Molten fluoride salts offer only marginal tritium breeding ratios. Ceramic breeders need a separate neutron multiplier, typically beryllium, which is expensive and resource-limited. No breeding blanket has ever been tested in a real fusion environment, and proving that one works reliably is one of ITER’s key missions.
Materials That Can Survive the Reactor
The inside of a fusion reactor is one of the harshest environments engineers have ever tried to build for. The plasma reaches temperatures above 100 million degrees, and while it doesn’t touch the walls directly, the neutrons it releases slam into surrounding materials with enormous energy (14 million electron volts each). Over time, these neutrons knock atoms out of position in the metal structure, a process measured in “displacements per atom” (dpa). A commercial fusion reactor would subject its walls to somewhere between a few tens and a hundred dpa over its lifetime.
Current experimental facilities have only tested materials up to about 1 to 10 dpa under fusion-relevant conditions. Beyond the neutron damage, the plasma-facing surfaces also endure direct erosion from hot particles escaping the plasma edge. Experiments show measurable surface erosion after relatively short exposure periods. No structural material has been proven to hold up under the combined assault of neutron bombardment, heat, and plasma erosion for the years a commercial plant would need to operate between maintenance shutdowns.
A Heat Flux With Nowhere to Go
In a tokamak, the exhaust heat from the plasma gets funneled to a narrow component at the bottom of the chamber called the divertor. The heat concentration there is extreme. Commercial reactor designs call for the divertor to handle at least 10 megawatts per square meter under normal operating conditions, with transient spikes up to 20 MW/m².
Current divertor designs struggle to meet even the baseline target. The copper alloy heat sinks used in today’s prototypes have a narrow safe operating temperature range, and testing shows that pipe-style cooling structures fail to keep materials within their design limits at 10 MW/m². A more advanced helium-cooled design narrowly met the target, but only under optimistic assumptions about material properties and perfectly steady heating. ITER’s tungsten divertor is designed to survive 5,000 thermal cycles at 10 MW/m² and 300 cycles at 20 MW/m², but a commercial plant running for decades would face orders of magnitude more cycles. Solving the divertor problem likely requires either new materials or entirely new exhaust geometries that haven’t been invented yet.
The Laser Approach Has Its Own Scaling Problem
The National Ignition Facility’s achievement used 192 laser beams to compress a tiny fuel pellet. That experiment fired once, and the lasers consumed roughly 300 megajoules of electricity to deliver about 2 megajoules to the target. For a laser-based (inertial confinement) fusion power plant, the driver would need to fire at a rate of about 10 times per second, with fuel pellets fed into the chamber at roughly 600 per minute. Each of those pellets would need to be manufactured to extraordinarily precise specifications, positioned perfectly, and compressed symmetrically enough to ignite.
NIF’s lasers were never designed for repetition; they need hours to cool between shots. A commercial inertial fusion plant would require entirely new laser technology that is both efficient enough to achieve net wall-plug energy gain and durable enough to fire millions of times without degrading. That technology does not currently exist.
Cost Projections Are Not Competitive
Even if every engineering problem were solved tomorrow, the economics of fusion remain challenging. Modeling suggests that early fusion plant designs would produce electricity at more than $150 per megawatt-hour. For context, fusion would need to hit roughly $80 to $100 per MWh to compete with other low-carbon options expected to be available after 2040. Variable renewables like wind and solar are already cheaper than that in many regions, and their costs continue to fall.
The first commercial fusion plant (called a “first of a kind” unit) would carry enormous one-time design and construction costs, likely experience significant downtime to address problems that only appear during real operation, and need an extended commissioning phase before it could generate revenue. Nuclear fission, which has a 70-year head start, still struggles with cost overruns in Western countries and typically lands above $80/MWh for new builds. Fusion would need to travel a similar learning curve, starting from a higher price point, with the added complexity of unproven fuel cycles and materials.
Timelines Keep Shifting
ITER, the project most likely to demonstrate sustained burning plasma, illustrates how timelines slip. In 2016, the ITER Council approved a schedule calling for first plasma in December 2025 and full deuterium-tritium operation in 2035. By mid-2024, the council received a revised baseline that pushes deuterium-deuterium operation (a stepping stone, not full fusion fuel) to 2035, with full-power operations further out still. The project has faced manufacturing defects in key components, pandemic delays, and budget overruns.
Private companies are moving faster in some respects. Commonwealth Fusion Systems is building SPARC using high-temperature superconducting magnets and leveraging AI-powered digital twins to accelerate design. But SPARC is a demonstration device, not a power plant. Its successor, called ARC, would be the actual power-producing machine, and no construction date has been set for it. Other private ventures have announced ambitious timelines, but none has yet demonstrated net energy gain from a device of their own design.
Regulation Is Still Being Written
Until recently, fusion existed in a regulatory gray area. The U.S. Nuclear Regulatory Commission is now proposing to regulate fusion machines under its existing framework for byproduct materials (the rules governing radioactive materials produced by accelerators and reactors) rather than applying the full weight of fission reactor licensing. This is a significant distinction: fission plant licensing can take a decade or more and costs hundreds of millions of dollars. Treating fusion devices more like particle accelerators would streamline the path considerably.
The proposed rule would classify fusion machines as a subset of particle accelerators but with their own specific requirements, separate from existing accelerator guidance. This framework is still in the rulemaking process, meaning no fusion company can yet receive a license to operate a power-producing reactor in the United States. Other countries are at various stages of developing their own regulatory approaches, adding uncertainty for companies planning to build internationally.