IMPERIALArticles

University spinouts

Universities should take less and license faster

Europe's universities should cut their stakes to published, standard terms, since spinout value has grown as stakes have shrunk.

· Universities

The painted front of the main building of ETH Zurich, seen from its entrance steps, with two globe lamps either side of the door
The main building of ETH Zurich, whose spin-offs raised CHF 540m in 2025. Photo: Leonhard Lenz, CC0, via Wikimedia Commons.

British university spinouts are now worth £49bn, which is 2.8 times what they were worth in 2020, according to the Royal Academy of Engineering on June 3rd 2026.1 Some 96 per cent of that value sits in deep tech, the sort of company that spends years in a laboratory before it earns a penny. Universities own a slice of nearly all of them. The slice is shrinking, and it is now about 16 per cent on average, the lowest in a decade.1

We think the direction is right and the pace is too slow. Europe's universities should publish their terms, license on standard paper and stop short of the point where founders start to resent them. A university is good at hosting a company and poor at picking which one will win. It should price itself accordingly.

The reform was not subtle. In the period leading up to 2023 the average stake ran at 22 to 25 per cent.1 The Independent Review of University Spinouts, published in November 2023, then recommended 10 to 25 per cent for companies built on heavy intellectual property and 10 per cent or less for software.2 Stakes fell and value rose. We do not claim the first caused the second; a good deal else has happened since 2020. But nobody has shown that taking a quarter of a company gets it built faster.

Some universities still ignore their own guidance. Software spinouts average around 17 per cent, against the recommended ceiling of 10 per cent or less.1 A software founder needs a lab least of all. A university that holds 17 per cent of a firm whose main equipment is a laptop is charging rent on a room the founder can leave.

Zurich shows what the alternative looks like. Oxford ranks first in Europe by spinout value and five of the top ten European universities are British, but Switzerland leads per head of population.1 ETH Zurich counted a record 46 recognised companies in 2025 (24 spin-offs and 22 start-ups), which raised CHF 540m across 41 disclosed rounds, 27 per cent more than the year before.3 A caution: the start-up label is new, so the count is not comparable with earlier years. In September 2025 ETH also introduced its first dedicated Equity and Licensing Policy, with an express licensing process on standard terms so that founders can start sooner.4 The law firm that advised on it gives no percentages. We would like to see them.

Our prescription is short. Publish the equity formula, by sector, on the university's website. Offer a standard licence that a founder can sign without hiring a lawyer for the negotiation. Treat the 10 per cent software ceiling as a ceiling. A licence should behave like a catalyst: it speeds the reaction and is not used up in it.

The strongest objection is money. Universities are short of it, a stake is how they pay for the next laboratory, and the returns are real. Oxford saw two exits above $1bn in 2025: OrganOx at about $1.5bn and Oxford Ionics at about $1.1bn.1 For illustration, 16 per cent of a $1.5bn sale is $240m, though we do not know what any university actually held.

The answer is in that very arithmetic. Spinout returns are lumpy: a handful of firms pay for the rest. A university's best hope is that the next OrganOx exists at all, and a founder who has read a fair term sheet is more likely to start one than a founder who has read an aggressive one. Squeezing the many weak firms for an extra few points yields little. Discouraging the one strong firm costs a great deal.

Universities have long insisted that results be replicated. They might now try replicating Zurich's paperwork.

The Imperial Review is an independent publication. Not affiliated with Imperial College London.