Roughly 77% of the money behind all of that is third-party funding.
Both facts come from the same report, and the second one determines how long the first one lasts.
What Gründungsradar 2025 shows German universities have built
The Stifterverband’s Gründungsradar is now in its sixth edition and is the most complete picture of entrepreneurship support at German higher education institutions. The 2025 edition reports on the 2023 data year - worth stating explicitly, because the report year and the data year are two years apart and the difference matters when comparing against more recent datasets.
The achievement side is genuinely impressive, and it should be stated before the critique. 2,927 foundings emerged from German universities, about 5% more than in the previous survey. Software led sectorally. Total budget for startup support reached roughly €218 million. Founding-relevant events reached around 400,000 people, 100,000 more than in the previous survey. 55% of institutions now cooperate with international investors, more than ten percentage points up on 2022. Co-creation projects with business and government more than doubled to 1,910.
The Deutscher Startup Monitor 2025 confirms the effect from the founder side: about 51% of surveyed founders received support from academic institutions and research facilities. Satisfaction is mixed but positive at 64%, with academic networks (nearly 63%) and knowledge transfer (54%) rated highest. Roughly 40% of German founders now consider Germany more attractive than the USA, up six percentage points year on year, and 61% rank Germany first within Europe.
This is a system that works. That is the point of what follows.
Why 77% third-party funding is a capability problem
Third-party funding is not inherently bad. Project money buys focus, brings in external standards and forces initiatives to justify themselves. A healthy support system uses it.
The problem is proportion. When 77% of the total budget is temporary and the institutions’ own contribution to the budget is a rounding error, the structure stops being a supplement and becomes the foundation. And 84.5% of surveyed universities are now explicitly calling on politics to establish permanent, dedicated funding for startup support. When five out of six institutions in a sector make the same demand, that is not lobbying. That is a structural report.
Here is what soft money actually does to entrepreneurship support, in mechanics rather than in complaint.
It puts staff on fixed-term contracts. The people who run these programmes are the ones who hold the mentor relationships, know which alumni will take a call, and remember which format failed three years ago and why. When their contracts expire, that knowledge leaves the building. Institutional memory in entrepreneurship support is almost entirely personal.
It shortens programme horizons to the funding period. A three-year grant produces three-year programmes, even when the capability being built takes longer.
It forces reporting cycles that reward countable short-term outputs - a topic that deserves its own article - over the slower work of building trust with founders.
And it makes the sector spend a meaningful share of its capacity applying for its own continued existence.
This matters more in entrepreneurship than in most disciplines, because the asset being built is relational. A mentor network is the accumulated result of years of individual relationships. An alumni pipeline only exists once a first cohort has been out long enough to come back. Founder trust in an institution is built one honest interaction at a time and destroyed by discontinuity.
The international comparison makes the compounding argument concrete, though it needs a careful caveat. In the 2026 European Deep Tech Report, ETH Zurich leads the world in alumni-founded, VC-backed deep tech startups since 2020 with 192, ahead of EPFL with 94, Cambridge with 67 and MIT with 35. ETH and EPFL together account for 289 VC-backed spinouts, 8 unicorns, roughly $24.2 billion in combined enterprise value and about $6.4 billion raised.
The caveat: these figures cover VC-backed deep tech specifically. MIT’s 35 is not a general statement about MIT’s entrepreneurial output and should never be quoted as one. What the comparison does show is what decades of continuous institutional commitment produce in one defined category - and neither ETH nor EPFL built that on a three-year grant.
What “institutional anchoring” would actually mean is unglamorous and specific: permanent positions rather than project posts, entrepreneurship professorships, credit-bearing venture creation formats inside degree structures, and a budget line rather than a funding application.
Entrepreneurship education is the only part of the ecosystem we ask to prove its existence every three years.
