Every fusion startup that has raised over $100M

The fusion money has decided. One company — Commonwealth Fusion Systems — has swallowed roughly a third of every private dollar ever committed to the sector. Its latest $1 billion round, closed this July, pushed its total haul to $3.94 billion. Everyone else is running for second place.

That concentration should tell you something about how investors actually see this technology. They are not betting on a portfolio of approaches. They are betting on a single tokamak design, built on high-temperature superconducting tape co-developed with MIT, led by a CEO who spent years inside MIT's Plasma Science and Fusion Center. The bet is that the doughnut works, that the magnets hold, and that the heat extraction scales.

The physics milestone that loosened the purse strings happened in December 2022. A Department of Energy lab fired lasers at a fuel pellet and got more fusion energy out than laser energy in. Scientific breakeven. Q > 1. The press releases called it historic. The fine print called it a long way from commercial breakeven — where the reaction pays for the entire facility, not just the lasers. That distinction matters. It is the difference between a science experiment and a power plant.

CFS says Sparc, its demonstration plant rising in Massachusetts, will hit Q > 1 sometime in 2027. The facility should operate by late 2026 or early 2027. That timeline is tight. Tokamaks have decades of operational history. High-temperature superconducting magnets do not. The tape-wound D-shaped coils CFS designed have never contained a burning plasma at power-plant relevance. Commissioning always finds problems that simulations miss. AI and better chips have sharpened those simulations. They have not eliminated the gap between model and machine.

The three drivers investors cite — powerful chips, sophisticated AI, high-temperature superconductors — are real advances. They enable control schemes that were impossible ten years ago. They allow faster iteration on designs. But they are tools, not solutions. The hard problems remain materials that survive 141-megavolt neutron flux, tritium breeding blankets that actually breed, and a supply chain for superconducting tape that currently barely exists. CFS will have to build or catalyze that supply chain itself.

Arc, the commercial follow-on, promises 400 megawatts later this decade. That number lands in utility-scale territory. It also lands in the territory where grid interconnection studies, regulatory approvals, and cooling water rights become engineering problems as hard as the physics. Fusion developers love to talk about the reactor. They talk less about the balance of plant — the turbines, the heat exchangers, the civil works that turn neutron flux into electrons on a wire. Those bits are boring. They also determine whether the economics work.

The capital concentration creates a systemic risk. If Sparc slips — if the magnets quench, if the plasma instabilities resist control, if the tape delaminates under cyclic loading — the entire private fusion narrative takes the hit. There is no diversified pipeline. The other startups that have cleared $100 million exist, but they live in CFS's shadow, raising at smaller scales, pursuing different configurations that now struggle to attract follow-on capital because the "winner" has already been anointed.

Markets hate a single point of failure. The fusion sector has built one anyway.

Skeptics have earned their skepticism. Fusion has been a decade away for seven decades. The difference this time is not that the physics finally works — the physics has always worked. The difference is that private capital has decided to underwrite the engineering at a scale that governments refused to sustain. That is a bet on execution, not a vote of confidence in inevitability.

Bob Mumgaard and his team know the difference. They have to deliver a machine that runs, not a simulation that converges. The $3.94 billion buys them time, talent, and titanium. It does not buy them Q > 1. That comes from plasma that behaves, magnets that hold, and hardware that survives. The clock on Sparc is ticking.