
The word “seed” used to mean one thing. Now it means two.
Crunchbase News published the numbers in March, and they are worth sitting with. US seed funding totaled $19.4 billion in 2025. But 51% of those dollars went to rounds of $10 million or more, up from a third in 2024. Seed rounds of $10 million and above climbed from 2% of all seed deals in 2018 to 9% in 2025, roughly one in ten seed deals over $200,000, numbering around 360. Meanwhile rounds of $5 million and under fell from 93% of deals in 2018 to 75% in 2025. (Crunchbase News, Mar 26, 2026; US seed data.)
Carta’s data says the same thing from the valuation side. The median seed post-money valuation has risen more than 40% since 2021, from $15.0 million to $21.6 million as of Q1 2026, across 20,000 equity rounds raised by US startups since 2021. And at the top end, Carta’s Q2 2026 data puts the 95th-percentile seed valuation at $200.4 million, up from $72.2 million a year earlier, an increase of about 178%. The median moved. The top moved three times as fast. (Carta, Round Benchmarking Tool launch, Jul 14, 2026; Carta Q2 2026 seed data as read by NextView Ventures.)
Dealroom gives you the cleanest single number for how unusual a big seed is. Aevius raised $22 million this week, and Dealroom puts that round in roughly the 99th percentile of all US health seed rounds on record, a sample of nearly 11,000 deals. (Dealroom, Nov 17, 2026.)
Two seed rounds announced Nov 17 show what the top half pays for.
Remanence, in Paris, raised a $30 million seed round co-led by EQT Ventures and FirstMark, with Zero Prime, Emblem, Drysdale, Global Founders Capital and Yellow, and angels including Datadog CEO Olivier Pomel and the Synthesia co-founders. It post-trains open-weight AI models into specialised enterprise systems, training models in simulated enterprise environments where AI agents practise tasks, make decisions and learn from mistakes. Its CEO, Pierre-Louis Cedoz, said: “We have seen first-hand how powerful post-training can be, as well as how difficult it remains for most companies to access. We started Remanence to close that gap and help businesses turn their own agent data into specialised models they control.” The founders are three former executives of the French AI lab H, with experience at Google DeepMind and Meta. The company claims its models can perform complex, multistep business tasks more reliably and potentially at a lower cost than general-purpose alternatives. That is the company’s claim.
Aevius, in Boston, raised $22 million led by Insight Partners. It is a Mass General Brigham spinout building AI models that make whole patient records computable. Its foundation model, APOLLO, makes complete patient records searchable, predictable and explainable in a single pass. The founders, Rowland Pettit, Ming Yang (Max) Lu, Faisal Mahmood and Long Phi Le, were part of the founding team behind Modella AI, which AstraZeneca acquired. “Our mission is to make medicine computable,” said Pettit, co-founder and CEO. The company was founded in 2026, and it has disclosed no customers or revenue yet. That matters, and I am stating it plainly: the round is a bet on a team that has done the hard part before, not on traction you can check today.
Neither case proves a round happened because of one proof point, or one résumé. I am not claiming that. I am saying the pattern is the same in both: investors at the top of the seed market are paying for a team that has already built in the same field, and for a specific, checkable thing that team can build next.
I wrote about the other side of this in IMPACT, in a chapter called “Live To See Your Success – Build a System.” Modeling successful people is the way to get amazing results in a short time, because successful people are different by definition. They are getting results that are not “normal.” Their results are extraordinary. As my friend and mentor Glenn Dietzel says, “Greater sameness will never get you better results.” That means you want to model not what they are doing the same, but what they are doing differently. (Book passage: Ken McArthur, IMPACT (2008), Chapter 12, “Live To See Your Success – Build a System” — https://kenmcarthur.com/amazon)
So here is what I would do if I were raising a seed round in the next 90 days. First, know which half of the market you are in. If your round is under $5 million, you are in the 75% of deals that share a quarter of the dollars, and your competition is not the $30 million seed. It is the other 75%. Second, size the round to the next proof point, not to the headline number. The round is the price. The proof is what the next round prices. Third, model what the successful teams do differently: one specific, checkable signal, and one trusted introduction to the investor who can say yes.
Seed has split in two. The founders who win in either half are the ones who know which half they are in.
If this was useful, forward it to one founder who’s raising. And if you are not subscribed to Raise Smarter yet, subscribe here: https://www.linkedin.com/newsletters/raise-smarter-7514657628552183808/
Founders: start your 21-day free trial of Founder Pro at https://connectlab.live/founders. Funds: see the founders raising now at https://connectlab.live/join/funds.
Connection is the currency of the 21st Century.
