“The rounds that were way too expensive back then are pretty much all way too cheap today.”

Filip Dames said this on the latest episode of my DEALMACHER Podcast. Filip is Founding Partner of Cherry Ventures, which he started in 2012 with Christian Meermann and which has backed companies like Flixbus, Auto1 and Proxima Fusion. We spent almost two hours on price sensitivity in funding rounds, deep tech, fundraising on both sides of the table and where Europe has to win. Six points from the conversation:

1. Discipline, and the exceptions that paid off

Cherry is disciplined on valuation. The target is around 15% ownership, sometimes 20 to 25%, often less. Filip would rather hold 8% of a company that goes through the roof than nothing at all.

But he also makes exceptions. In the episode he described the moment: an extraordinary team, a price that is “way too expensive, but there’s something we feel, and we absolutely want to be in.” The times Cherry stretched like that, he says, it was mostly the right decision.

His read of today’s market explains why. Most of the money flows into a small group of companies. If you are in that group, the price barely matters. If you are not, even a very good company gets surprisingly little.

2. Depth before gut

The Proxima Fusion deal shows how Filip works. He wanted a fusion investment for the fund, read everything he could find and spoke to almost every team.

Magnetic fusion reactors hold a super-hot plasma in place with magnets. The classic design, the tokamak, is a ring-shaped chamber that is relatively easy to build. Its weakness: the plasma is hard to keep stable for long. Proxima builds a stellarator instead. Its magnets are twisted and curved with extreme precision, which keeps the plasma stable but makes the machine much harder to construct.

His thesis: with AI, the team can run thousands of construction simulations a day and calculate the exact curvature of every magnet. That makes the construction problem easier to solve than the stability problem. Then he met the team, and the decision was clear.

Filip codes himself these days. Asked to choose between deep technical understanding and being a great fundraiser, he clearly picked the first. Cherry does twelve, maybe fifteen investments a year, three to four per partner. That leaves the time to go this deep.

3. How founders raise in this market

Filip’s sequence for founders starts with the story. He told me European founders are still far behind Americans here: great companies, great products, and they still can’t sell them. I see the same with deep-tech founders I advise. The most important fact often sits in a subordinate clause.

Once the story works, “it’s a numbers game.” That means the right partner inside the fund, not just the fund. A warm route in. A way to stand out from hundreds of pitches. Scarcity and momentum.

And one line every founder should remember: “You have not been preempted until you have been preempted.” Until a term sheet arrives, a fund that “might preempt” is having a coffee chat. Some are collecting information. Some may even be invested in your competitor.

4. San Francisco, not New York

For AI companies, Filip’s answer is clear. Founders need to spend real time in San Francisco. New York feels more European, and six hours of time difference beats nine. “It’s the easy choice. It’s the comfortable choice. It’s not the right choice.”

There are exceptions. Fintech and retail can work in New York, and so can a company whose customers sit there, like the law firms Legora sells to.

Two more points from the episode. Part of the founding team has to move; a country manager steered from Germany is something he has never seen work. And most founders go too late. He is not dogmatic about timing, though. A robotics or industrial company can spend its first two years winning Europe, where many of its customers are.

5. A fund is a product, and LPs check it every three years

“For me, Cherry isn’t a fund. It’s a company, and a company has a product.” Filip expects European venture to consolidate. Funds between the mega-platforms and the solo GPs need a brutal differentiation, because the best founders hold ten or fifteen term sheets and ask: “Why should I take money from you if I have 14 other options on the table?”

The same logic applies upstream. Most GPs hate LP fundraising, he says. He enjoys it. Every three years the fund has to ask: did we perform, and did we earn more capital? Without that moment, a fund is not really accountable to anyone. The early funds meant knocking on doors, 100 meetings. The first institutional fund was 150 million euros. Some LPs Cherry met for Fund 2 in 2016 invested for the first time in Fund 4 in 2022.

6. What AI doesn’t commoditize

Cherry’s investment memos used to run about eight pages. With AI in the diligence process, they grew to 40 or 50. Filip’s message to his team: if you cannot explain in five to seven pages why you want to make an investment, you probably don’t want to make it.

The intelligence layer is becoming a commodity, he said in the episode. What isn’t: “judgment, foresight, empathy.”

His bet for Europe is just as clear: AI, energy, defense, space. If those companies are not built here, Europe will pay rent to the US and China to use theirs.

You can listen to the full conversation with Filip, including the question of whether Berlin is still the right place to build, at www.id.vc/dealmacher-podcast.