Last week, I was talking with a partner at a tier-one fund in San Francisco. He told me something surprising: the average age of founders in their pipeline has dropped to 25, and the average valuation on those deals runs $30 to $40 million. They focus more than ever on founders who are exceptional at hiring. None of this would have made sense five years ago, except for focusing on founders who can build good teams.
A few thoughts on what’s happening in pre-seed right now:
The accelerator culture that used to live on a few blocks in San Francisco has spread across the whole industry. Twenty-two-year-olds are watching their cohort raise at double their valuation and reading it as a verdict on themselves. I know founders struggling under the weight of a number some other 22-year-old got, and they didn't.
AI has created much opportunity for younger founders. The ones with the steepest learning curves are absorbing new tools fastest, and at that age you can throw every hour you have at one problem without anything else competing for it. Mercor is the case everyone points to. But the base rate hasn't moved. The large majority of big outcomes still won't come from 18- to 22-year-olds, and treating every sharp 22-year-old like the next Mercor is how a market talks itself into something that isn't true. At least not yet.
The data cuts the other way more often than the pitch decks admit. The average founder is getting younger, but the average unicorn founder is getting older: up to 40 by 2020, even as the broader founder population drifted down toward 35. Age isn't the edge. Speed is, and speed doesn't only belong to the young anymore.
That's what's actually new about this moment: Thirty-six percent of startups founded on Carta in 2025 were solo-founded, double the rate from a decade ago, and AI is a big part of why: one person can now do work that used to take a co-founder and two hires. But only 17% of VC-funded startups were solo-founded last year. Founders are proving they can build alone faster than investors are proving they'll fund them alone.
Finding a serious company means digging through more noise than I've seen in eight years of doing this. The gap between good and bad is widening. A lot of tourists are entering the arena to build AI slop copycat companies with no underlying reason why the company should exist.
Every third deck landing in my inbox reads as if it came out of the same LLM, because increasingly it did. Founders are running their positioning through the same tools and getting back the same deck with cringey transitory copy.
Physical AI is having the same moment right now. It showed up in San Francisco a year and a half ago, and most of the capital chasing it is going to assemblers, companies stacking components into a finished product. That's backwards. The durable value sits one layer down, with the component manufacturers building the gears and gadgets nobody else can replicate. Assemblers compete on speed to market. Component makers own the bottleneck.
Data will play out the same way. For most applications, incumbents already sit on the data that would make a challenger's model good, and no amount of clever prompting closes that gap. Tech-enabled services still have openings in industries nobody's paying attention to. Legal isn't one of them. That market has already been decided.
The venture capital industry that used to move on a yearly clock is now moving weekly, and most of the capital chasing it hasn't reset its own clock to match.
AI as a category is cooling off. The LLMs are rushing to go public. Semiconductor stocks shed $1T+ in market value in July as Wall Street questioned AI capex sustainability. Capital deployment in H1 2026 peaked, and it seems we are starting to fall back down the mountain. Tokens have been underpriced to accelerate adoption, so when they rise, the industry will consolidate. Burning down the forest and seeing who comes out standing is generally a positive event.
Five years ago, a fund picked a thesis and lived with it for twelve to twenty-four months. Now every quarter is a sprint into whatever the last few weeks produced. It’s good to participate in the hype trends, but with discipline and patience. We can afford to be wrong on a check. We can't afford to be absent from the shift. People are deploying capital into categories they haven't taken the time to understand, because the category will have moved again before they finish learning it.
The money is consolidating into fewer hands. The top 5% of seed valuations are 10x where they sat a decade ago, according to Carta. The number of companies getting funded is shrinking as dollars pile into a smaller list of names. Pedigree still drives capital more than the results justify, but pedigree doesn’t equal outlier returns; if anything, it’s quite uncorrelated. Talent has consolidated the same way. San Francisco now has 10x the tech talent New York does, and New York sits closer to any other hub in the country than it does to San Francisco.
That consolidation shows up in the money too. San Francisco's share of US venture dollars sat around 41% in 2023, already a multiyear high at the time. It jumped to 57% in 2024 ($90B of the $178B raised nationally), then 60% in 2025 ($122B), then over 80% of Q1 2026 dollars and 68% of Q2, a blended low-to-mid 70s for the first half of the year. Three years ago, San Francisco taking less than half the country's venture dollars would have been the story. Now the surprise would be if it didn't.
Consolidated capital and inflated round sizes are building the biggest opportunity in the market.
As valuations keep outrunning the size of the rounds funding them, downstream investors are running out of clean paths to buy equity, and that gap is going to widen. Early-stage, small-check writers are sitting on the equity those investors are buying, providing early liquidity for the nimble early stage funds.
The next wave of outcomes won't be dominated by home runs. It'll be built on doubles and triples, companies that stay small to medium and still hand their founders eight figures. The middle class of VC is shrinking. For most founders, that's already the goal, a repeatable outcome the industry has spent the last decade discounting because it doesn't make as good of a headline. The barbell is widening where you can make decent multiples on Seed-strapped companies, or you pay elevated prices for $100B outcomes, not the $1B ones that used to count.
Pre-seed in 2026 is louder, more consolidated, and more expensive at the top than ever. The opportunity isn't in trying to outrun that noise. I believe it's being able to lean in and look past the noise at the same time; finding the outsider founder flying under the radar while also supporting companies being built on the arc.
