Every Fusion Architecture Chooses Its Bottleneck

The investment case for fusion depends on the cost of turning scientific progress into dependable electricity

Fusion's commercial winner must do more than produce an impressive experiment. It must reach useful power with a fundable construction budget, operate reliably, and leave investors with enough ownership to earn a return. Reactor architecture matters because it shapes all three. Our investment thesis favors compact, maintainable plants that reduce the capital needed to prove the next stage. The evidence supports that direction, but does not yet establish a winning architecture.

Start with the electricity that can be sold

Investors should clarify what "gain" measures before comparing any technology. Plasma gain compares fusion output with energy supplied to heat the plasma; target gain compares output with energy delivered to the target. Neither includes every loss between the electricity supply and the reaction. For this article, engineering gain means gross electrical output divided by all electricity consumed by the plant over the same operating cycle. Net electricity is the surplus left after that consumption.

NIF illustrates the difference. In April 2025, the National Ignition Facility produced 8.6 megajoules (MJ) of fusion energy from 2.08 MJ of laser energy delivered to its target, a target gain of 4.13. Lawrence Livermore National Laboratory (LLNL) separately describes about 400 MJ of stored electrical energy delivered to the laser flashlamps per shot. Using that approximate input, 8.6 ÷ 400 is only 2.15%-a fusion-energy return on stored electrical input, not an electrical engineering gain. [1, 2]

Assume, purely for illustration, that 40% of the fusion energy could become electricity. The resulting electrical-return proxy is (8.6 × 40%) ÷ 400 = 0.0086, or 0.86%, before charging losses and other facility loads. The 400 MJ figure is not a complete, shot-specific electricity audit, and NIF has no electricity-generating system. This calculation therefore illustrates the gap; it does not report a measured engineering gain. [1, 2]

The commercial task is to close that gap while operating repeatedly. LLNL's development requirements for future inertial-fusion laser drivers include at least 10% wall-plug efficiency and repetition rates around ten shots per second. NIF's scientific achievement is an important input to that effort, rather than a demonstration of its economics. [3]

Four approaches and the costs they inherit

Tokamaks and stellarators both use magnetic confinement. Inertial fusion compresses small fuel targets, while magnetized target fusion combines magnetization with compression. The categories below simplify the landscape; the examples have different objectives and levels of maturity. [3-7]

Company examples describe design intent, not validated plant performance. Thea uses planar coil arrays to address conventional stellarator manufacturing complexity. General Fusion's proposed commercial liquid-wall cycle should not be confused with its LM26 demonstration. [6, 7]

Does commercial fusion need to become smaller and simpler

These bottlenecks lead to a broader investment question: can the next design reduce the cost of proving and deploying fusion? Compactness is attractive when it means less construction, a smaller funding commitment and faster learning. It is a design attribute, not a separate fusion category.

There is a credible technical basis for this thesis. The peer-reviewed ARC study describes how stronger superconducting magnets could enable a smaller tokamak and reduce capital cost. The MANTA design study projects a compact pilot plant with about 90 MW of net electricity and a $3.4 billion overnight construction cost. These are modeled designs, not construction bids or operating results. [8, 9]

Smaller also concentrates heat and leaves less room for shielding and maintenance. ARC heat-exhaust research treats those constraints explicitly. Our inference is therefore conditional: compact designs have an attractive commercialization path if they preserve component life and service access. The literature supports pursuing this advantage; it does not justify declaring compact fusion the most likely scientific winner. [10]

Large plants create a financing bottleneck

The capital bill extends well beyond the reactor: power supplies, buildings, cooling, fuel handling, shielding, maintenance equipment and, for thermal designs, turbines. Larger plants may spread fixed costs over more electricity, but they also require larger commitments before operating evidence is available. A cheaper reactor core is valuable only if it reduces the installed cost and operating burden of the whole plant.

The National Academies' 2021 report placed an indicative $5-6 billion ceiling on overnight construction cost for a commercially viable first fusion plant in the U.S. market of that time, and recommended keeping a pilot below that range. This was an affordability assessment, not a current price quote. "Overnight" also excludes financing costs accumulated during construction. Delays can therefore increase the funding requirement even before equipment costs change. [11]

Consider an illustrative 1 GW net-electric plant costing $5,000 per kW: the construction bill is $5 billion. With an assumed 8% annual capital-recovery charge and 90% capacity factor, capital alone contributes about $51/MWh. At 60% capacity factor, that rises to $76/MWh, before operating costs, replacements and fuel. These are scenario assumptions, not a fusion cost forecast. They show why maintenance downtime can undo an apparently attractive construction budget.

A compact plant need not have the lowest cost per kilowatt to be a better first investment. A smaller absolute check can make a demonstration easier to fund and a failed iteration less costly. But commercial deployment ultimately requires both affordable total capital and competitive electricity. The next architecture must demonstrate net output, maintainable components and repeatable manufacturing in one integrated system.

Valuation determines dilution rather than technical success

Large funding rounds can buy time and engineering capacity, but the valuation itself is not a probability of success. In June 2026, Helion announced $465 million in funding at a $15.5 billion post-money valuation. That is a financing outcome, not independent validation of commercial net power. [12]

For a simple primary equity round, new investors' ownership equals capital raised divided by post-money valuation. A company raising $500 million at a $5 billion post-money valuation sells 10%; at $10 billion, it sells 5%. Higher valuations let founders and existing investors fund expensive development while giving up less ownership. This arithmetic excludes option-pool changes, convertibles and other deal terms.

For the incoming investor, the higher price raises the return hurdle. A 5% stake bought for $500 million needs a $30 billion equity exit to return 3× before further dilution, assuming pro-rata proceeds. If future financing halves that ownership, the required exit becomes $60 billion. Technical success can therefore coexist with a disappointing investment return.

The investment decision should connect each round to the uncertainty it removes and the capital still required afterward. Every fusion architecture chooses its bottleneck. The strongest commercial proposition makes that bottleneck affordable to solve-and leaves a credible path from a working machine to electricity customers will pay for.