56 Fusion Startups Say They Need $77 Billion. The Math Says 40 Won't Survive to Spend It.

Seventy-seven billion dollars. That is how much the world's 56 private fusion companies say they still need to reach pilot-plant stage, according to the Fusion Industry Association's 2026 Global Report, released this week. They have raised $14.2 billion to date, which means the gap between ambition and capital is not a rounding error but a factor of 5.4, the kind of chasm that would take 17 years of uninterrupted funding at the current record-setting pace of $4.48 billion per year to close, assuming zero operating burn. Nobody in fusion operates at zero burn.

That $4.48 billion figure, a 69% jump over last year and the highest annual investment the FIA has ever recorded, is being celebrated across the industry. General Fusion went public on Nasdaq under the ticker GFUZ on July 13, becoming the first publicly listed fusion company. Commonwealth Fusion Systems has its SPARC tokamak 65% complete in Devens, Massachusetts. Helion Energy broke ground on its commercial-scale Orion plant in Washington State. But dig beneath those headlines and a harder arithmetic emerges: one that divides the 56 companies into a handful of probable survivors and a much larger group running out of runway.

Four Unicorns, Fifty-Two Followers

The F4E Fusion Observatory's November 2025 report identified 77 fusion companies globally and tracked cumulative funding of €13 billion ($14.3 billion), but four companies hold the majority of that capital, a concentration pattern that determines who lives and who dies. Commonwealth Fusion Systems leads at €2.6 billion, followed by China Fusion Energy and NEO Fusion at €1.9 billion each, and TAE Technologies at €1.3 billion. Combined: €7.7 billion, roughly 59% of all private fusion capital ever deployed.

Strip those four out and the remaining companies split €5.3 billion among themselves, an average of €73 million each, which sounds like real money until you consider that building even a modest demonstration tokamak runs into the hundreds of millions before anyone pours the concrete pad for the reactor hall.

Run the same concentration logic forward against annual funding rates and the picture sharpens further. The FIA's 2026 report shows that the big raises dominate each year: CFS closed $863 million in August 2025, Pacific Fusion raised a $900 million Series A, Helion pulled in $425 million, and those three rounds alone consumed half the annual total. If the top eight companies absorb 75% of the $4.48 billion annual inflow, the remaining 48 companies share $1.12 billion per year, which works out to $23.3 million per company.

At $23.3 million per year and a median self-reported need of $700 million, each mid-tier fusion company requires 30 years of funding at current allocation rates to reach pilot-plant stage. Thirty years. The entire modern fusion startup ecosystem is younger than that timeline.

The Burn Arithmetic

Not every company burns at the same rate, and not every company needs $700 million. Some are computational modeling shops running lean on $5-10 million per year. Others, like Realta Fusion in Madison, Wisconsin, have raised about $60 million and are building hardware, including the first private demonstration of direct plasma-to-electricity conversion. But the physics of fusion reactors imposes a capital floor that software companies never face: superconducting magnets cost millions per unit, tritium breeding blankets require specialized materials from a handful of suppliers worldwide, and vacuum vessels weigh tens of tons while demanding machining tolerances measured in fractions of a millimeter.

General Fusion, founded in 2002 and backed by Jeff Bezos since 2011, illustrates what happens when capital pressure meets physics-constrained burn rates. The company laid off 25% of its staff last year before finding a lifeline through its SPAC merger with Spring Valley Acquisition Corp. III. The $150 million raised in that transaction funds operations through 2028, when CEO Greg Twinney told Barron's the company expects to have hit enough scientific milestones with its Lawson Machine 26 that raising additional money will get easier, a reasonable bet and also an admission that a 24-year-old company with 200,000 plasma experiments behind it still cannot self-fund.

Realta's situation is different in degree but not in kind: its next machine, Hammir, will be five times the length of the current WHAM device, with construction expected to finish by 2029 and commercial fusion targeted for the mid-2030s, yet the company has raised approximately $60 million total, most of which Hammir alone will consume.

Consolidation Math

Here is the calculation nobody in the industry is running publicly, laid out in full so you can check it.

Start with the inputs: 56 companies, a median remaining capital need of $700 million each, annual industry investment of $4.48 billion, top-eight concentration at roughly 75%, which leaves residual capital of $1.12 billion per year for the other 48 companies, average mid-tier annual burn between $20 million and $50 million, and average mid-tier cash on hand between $50 million and $100 million based on the F4E average of €73 million for non-unicorn companies.

Tier	Companies	Avg. Raised	Annual Funding Share	Years to $700M Goal
Unicorns (>$1B)	4	~$2.1B	~$840M each	On track / near-funded
Leaders ($200M-$1B)	~6	~$400M	~$200M each	2-3 years
Mid-tier ($50M-$200M)	~15	~$100M	~$35M each	17 years
Early-stage (<$50M)	~31	~$20M	~$13M each	52 years

Those mid-tier and early-stage companies are not going to operate for 17 to 52 years at current funding rates. They are going to run out of cash in two to four years, at which point they face three options: raise a breakout round that catapults them into the leader tier, get acquired by a larger competitor for their IP and engineering talent, or shut down.

My estimate: by 2031, at least 35 of the current 56 companies will have been acquired, merged, or dissolved. The industry will converge to 12-15 active players, with five to seven having funded a prototype reactor. This is not a prediction of failure for fusion itself. It is a prediction that the capital structure of a hardware-intensive, physics-constrained industry cannot support 56 parallel bets when each bet costs three-quarters of a billion dollars.

Where the Winners Are

CFS is the clearest frontrunner. SPARC, a compact tokamak using high-temperature superconducting magnets developed with MIT's Plasma Science and Fusion Center, is targeting first plasma in late 2026 and a Q>1 demonstration (more energy out of the fusion reaction than put in) by 2027. Power purchase agreements with Google and Eni are already signed for its ARC commercial reactor, a 400-megawatt plant planned for Chesterfield County, Virginia, in the early 2030s. Eni's deal alone is worth over $1 billion.

Neither party disclosed exact pricing or duration. But consider a rough envelope calculation. If Eni is purchasing half the plant's output (200 MW) over a 20-year contract at an 80% capacity factor, total energy delivered would be approximately 28 million MWh. A billion dollars divided by 28 million MWh yields an implied PPA price around $36 per MWh. For context, the World Nuclear Industry Status Report 2025 puts new-build nuclear fission at $123-180/MWh and utility-scale solar at $58/MWh. If the Eni number is anywhere near $36/MWh, CFS is promising baseload electricity cheaper than solar farms produce intermittently. Bold pricing or desperation pricing, depending on whether SPARC works.

Helion is taking a different technical approach and a different commercial strategy. Its binding power purchase agreement with Microsoft commits 50 megawatts by 2028 with financial penalties for non-delivery, underwritten by Constellation Energy as transmission partner. Helion's Orion facility in Malaga, Washington, is the hardware backing that promise. Its fusion approach, pulsed field-reversed configurations instead of steady-state tokamaks, is less proven than CFS's but potentially cheaper to scale, since it avoids the massive superconducting magnet assemblies that dominate tokamak costs.

The ITER Comparison Nobody Wants to Make

ITER, the 35-nation fusion megaproject in southern France, was supposed to cost €5 billion and achieve first plasma in 2016, but the current budget exceeds €25 billion, first plasma is now expected in 2034, and energy-producing fusion reactions, limited to short bursts, will not happen until 2039, at which point the reactor will still never generate a single watt of grid electricity, because it was designed purely as an experiment.

Private companies are lapping ITER on timeline if not yet on physics. CFS broke ground on its facility less than four years ago and is now assembling the tokamak itself. Consider the per-year spending comparison: ITER has burned roughly €760 million per year across 33 years of development for a reactor that will not produce electricity, while CFS has spent approximately $2.8 billion total across its entire existence and is building both a demonstration reactor (SPARC) and a commercial design (ARC) simultaneously.

ITER's defenders note, correctly, that CFS has not yet achieved Q>1 and that SPARC is designed for a much lower Q (target >2) than ITER's Q=10 goal, because scale matters in fusion and the physics of plasma confinement changes substantially as reactor volume grows. But the commercial question is not "which machine achieves the highest Q?" It is "which approach puts power on a grid first?" Private fusion's implied answer is faster, smaller, and cheaper, even if less scientifically ambitious.

The Strongest Case Against

In April 2026, researchers at ETH Zurich published a paper in Nature Energy titled "Fusion power unlikely to become competitive." Their argument is devastating in its simplicity. They interviewed 28 fusion experts across public and private sectors, assessed the technology characteristics of dominant reactor designs (magnetic and laser inertial), and matched those characteristics to experience rates observed historically in technologies of similar complexity. Large unit size, extreme design complexity, moderate-to-high customization: these traits correspond to experience rates of 2-8%, meaning costs fall by only 2-8% with each doubling of cumulative production.

At those learning rates, even starting from the optimistic end of first-of-a-kind cost estimates ($1,400 per kW), fusion requires an unrealistic number of cumulative doublings to reach cost parity with solar ($58/MWh) or onshore wind ($61/MWh). At the pessimistic end ($43,000 per kW for a 400 MW plant, implying $17.2 billion per reactor), the math is simply absurd.

Tang, Noll, Panda, and Schmidt recommend that policymakers "re-evaluate public funding" for fusion and redirect resources toward designs with more promising characteristics. Their logic is internally consistent and based on a methodology that has produced accurate predictions for other energy technologies.

Fusion advocates have two responses, one weak and one strong. The weak response is that 28 expert interviews cannot capture the diversity of approaches across 56 companies, which is fair as far as it goes, but the paper's conclusions apply specifically to magnetic confinement (48% of firms) and inertial confinement (21%), which together represent 69% of the industry and are the approaches that every major PPA-signing company is pursuing. The strong response is that high-temperature superconducting magnets represent a technology discontinuity that historical experience rates cannot capture, because CFS's magnets produce magnetic fields of 20 tesla in a compact form factor while ITER's older low-temperature superconductors top out at around 12 tesla and require a reactor roughly 10 times the volume, which means that if the HTS magnets are genuinely a step-change rather than an incremental improvement, the entire cost trajectory shifts and the ETH model, built on historical analogies to fundamentally different machines, misses the curve entirely.

Whether HTS magnets constitute such a discontinuity is the $77 billion question. SPARC's performance in 2027 will answer it.

What We Don't Know

This analysis carries several acknowledged blind spots. The $77 billion figure comes from self-reported survey responses in which companies estimated their own additional funding needs; such figures carry obvious incentive distortions in both directions (overestimating to signal ambition, underestimating to avoid scaring investors). Capital concentration percentages are derived by cross-referencing two different datasets (FIA's annual survey of 56 companies and F4E's observatory tracking 77 companies) that use different counting methodologies and reporting periods. The PPA implied price of $36/MWh for the CFS-Eni deal is illustrative, not factual, because neither party has disclosed the contract's MW allocation, duration, escalation terms, or force majeure provisions. China's €4.4 billion in fusion investment is overwhelmingly state-directed through entities like China Fusion Energy and NEO Fusion, which operate under fundamentally different capital logic than VC-backed Western startups, making a unified "industry consolidation" prediction imprecise for the global picture. Finally, the $4.48 billion annual funding rate used as the denominator in survival calculations could change dramatically; if AI-driven data center power demand continues to surge, fusion investment could double or triple within two years, altering every timeline in this article.

What You Can Do

If you are an investor considering GFUZ (General Fusion's Nasdaq debut), understand what you are buying: a pre-revenue company with $150 million in cash, a 2035 commercial target, and magnetized target fusion technology that uses steam-driven pistons instead of the superconducting magnets favored by larger competitors. Shares soared 25% on their second trading day, momentum that reflects retail enthusiasm for a category, not a fundamental repricing of General Fusion's specific technology risk. Watch for their Lawson Machine 26 milestones through 2028; those results will determine whether the company raises its next round at a premium or a discount.

If you work in energy procurement for a data center, utility, or large industrial consumer, fusion PPAs are entering the market. Both CFS and Helion have signed binding or near-binding agreements with Microsoft, Google, and Eni. These contracts price power delivery in the early 2030s, meaning procurement teams need to evaluate them against a portfolio that includes fission SMRs ($60-90/MWh projected), geothermal ($45-75/MWh), and long-duration storage. Fusion PPAs likely carry implicit "hell or high water" termination penalties that compensate for technology risk, so read the fine print.

If you are a policymaker evaluating fusion funding, the ETH Zurich Nature Energy paper deserves a careful read, but so does the FIA's data on private capital velocity. Public funding for fusion ($800 million in the latest year, an 84% increase) is a small fraction of private investment, and its highest-value use may not be funding individual reactor designs but building shared infrastructure: tritium supply chains, materials qualification programs, and regulatory frameworks that benefit the entire industry.

The Bottom Line

Fusion is attracting more money faster than ever before, and that money is concentrating into fewer hands faster than ever before. Both trends are signs of an industry maturing, not one failing. The $77 billion gap is real but misleading; not all 56 companies will survive or need to. What matters is whether the five to seven best-funded companies can convert capital into working reactors before the patience of their investors, their PPA counterparties, and the planet's carbon budget runs out. CFS's SPARC, expected to demonstrate net energy gain by 2027, is the single most consequential test. If it works, the consolidation this article predicts becomes a natural selection event that strengthens the survivors. If it does not, the ETH Zurich researchers will have been right, and $14.2 billion will join ITER's €25 billion as evidence that the sun's trick remains stubbornly inimitable on Earth.