AI Ate 82.7% of US Venture Dollars, and the Term Sheets Changed

U.S. venture deal value hit $515.8 billion through September, and 82.7% of it went to AI — one third of the world's venture dollars to OpenAI and Anthropic alone. Those figures anchor the argument Louis Lehot, partner at Foley & Lardner LLP, took to the main stage at the One Way Summit in San Francisco, where he sat with Immad Akhund of Mercury, Nicolas Dessaigne of Y Combinator and Suzanne DiBianca of Salesforce. Scott McGrew of NBC Bay Area moderated the panel, titled "Investing in the Impossible: The 2027 Playbook."

McGrew opened with a thesis: AI has made software cheap to build, so capital may shift to what AI cannot code into existence — fusion, robotics, quantum, energy, climate. Lehot called half of that right. Fusion raised a record $4 billion this year and robotics more than $74 billion. Funds that wrote billion-dollar checks for model labs now write $500 million checks for a tokamak.

What changed since 2021?

Asked what would have looked like science fiction five years ago, Lehot pointed to physical AI: hardware and robots taking over dirty, dull or dangerous tasks. Dessaigne quantified the shift — hardware has gone from 5% to 25% of the Y Combinator portfolio — while cautioning that venture still prefers software for its capital efficiency.

Lehot's lawyerly framing: "The reactor is not the science fiction. The cap table is." A founder with no product, no revenue and a fifteen-year path to a power plant once raised a seed round. This year, that founder raised hundreds of millions at Series A. Commonwealth Fusion Systems closed a $1 billion round before building its plant, with a venture fund, a sovereign wealth fund and a hyperscaler signing on day one, and the power purchase agreement signed before the physics is proven.

Is AI breaking the venture model?

Lehot's answer: AI is tearing the model down the middle. Software companies that needed a $5 million seed now need a credit card and three people, and fight over control, not price. At the other edge, model labs, data centers, fusion and robotics rounds read like project finance: milestone tranches, compute commitments, strategics signing commercial contracts the same day as equity.

What is dying is the middle — the $20 to $50 million Series A and B. Non-AI Series A companies get median valuations of about $42 million; AI companies at the same stage get $78 million.

What about subscription pricing?

Lehot said subscription business models are on life support. Akhund pushed back hard, saying seat-based subscription pricing still dominates what he sees across thousands of Mercury startup accounts. Dessaigne noted OpenAI and Claude run big subscription businesses that are not necessarily profitable. DiBianca backs consumption and outcome-based models, and her advice to founders was blunt: focus on EBITDA.

Dessaigne settled it: "At YC you do not bet on the business model. You bet on the founders." Lehot's landing point: systems of record remain the source of value, and the data rights clause now matters more than the pricing schedule.

Who wins, incumbents or startups?

Corporate investors account for roughly 88% of U.S. AI venture deal value this year — dollars, not deals. Lehot's advice to founders: take the money, but read the side letter.

How do you fund 20-year technology?

Not with ten-year funds. Lehot described four patterns in current fusion, quantum and robotics financings: time-horizon-layered cap tables; milestone-tranched rounds that read like construction loans; government money as a real tranche; and manufactured liquidity via continuation funds, structured secondaries and public listings a decade from revenue — "refinancings dressed up as exits."

Agents and money

Akhund said agents today read, recommend and draft, but humans approve transfers. His safeguards: spending limits, approval thresholds, the ability to stop an agent cold. Lehot heard a liability gap: most contracts do not say who eats the loss when an agent errs. "By 2027 they will."

DiBianca highlighted Salesforce's 1% equity pledge at founding, copied and exceeded by Google and Okta. Lehot cited Anthropic's Long-Term Benefit Trust as "a legal commitment with teeth."

Anthropic's IPO, discussed at around $2 trillion, would return real cash to early funds that have barely distributed a dollar since 2020-2021. "It would not fix the exit drought. It would be the first rain in three years," Lehot said.

His 2027 prediction: 992 active unicorns worth $5.7 trillion on paper, against 18 IPOs in a quarter, means a quiet recap — down rounds, secondaries at 40 cents on the dollar, pay-to-play financings — orderly, and beside AI winners still raising at record valuations. The evidence, as always, will be in the term sheet.