Seven billion dollars is a lot of money to spend on something that has been fifteen years away for seventy years. TechCrunch's deep dive into every fusion startup that has raised over $100M maps a funding landscape that looks, structurally, like every other deep-tech gold rush: capital concentrating around a handful of players, timelines perpetually optimistic, and a media cycle that treats press releases as proof of progress. The piece lands the same week that Joshua Kushner publicly warned that AI euphoria is weakening investment discipline across Silicon Valley. Fusion is the older version of the same story.
The Concentration Problem in Deep-Tech Funding
Tim De Chant's reporting reveals that the $7.1 billion raised by fusion startups has flowed overwhelmingly to a small cluster of companies. Commonwealth Fusion Systems and TAE Technologies have captured the lion's share, leaving a long tail of smaller bets. This is not unusual for capital-intensive frontier technology, but it creates a particular risk: when the top three players define the narrative, the broader field loses the diversity of approach that actually produces breakthroughs. A 2026 arXiv paper on AI reasoning by Rachel Lawrence and Jacqueline Maasch argues that autonomous reasoning is learnable and rule-based rather than emergent, a direct challenge to the field's dominant assumptions. Fusion faces a structurally identical epistemological problem: the most-funded approach (tokamak confinement) may not be the right approach, but it is the approach that attracts capital because it is legible to investors.
Kushner's Warning Is Actually About Fusion Too
Kushner's letter to Thrive investors, reported by TechCrunch, warned that excitement is weakening due diligence. He was talking about AI, but the sentence applies verbatim to fusion. The difference is timeline: AI produces revenue signals quickly enough that the market corrects fast. Fusion corrects slowly, expensively, and publicly. The investors who funded fusion in 2021 will not know if they were right until the 2030s. That is an unusually long period during which discipline-weakening euphoria can compound. TurboFund's breakdown of investor research mistakes identifies exactly this pattern: founders in long-cycle industries exploit the legibility gap between what is verifiable now and what only becomes knowable later. Fusion is the extreme case of that gap.