Pre-seed AI rounds in 2026 look very different from 2023. Here's our data from 200+ deals we've seen this year.

Round sizes have grown

Median pre-seed round for AI is now $2.1M (up from $1.4M in 2024). Top-decile rounds clear $5M. Compute is no longer cheap — founders are raising bigger pre-seed checks to fund 12–18 months of model training and team.

Valuations are bifurcating

We're seeing a barbell. Ex-FAANG founders with technical IP are pricing at $25–$40M post. Everyone else clusters around $8–$15M. The middle is hollowing out. If you don't have a sharp differentiator, expect the lower end.

What investors actually fund pre-seed

We're funding teams with: (1) deep technical credentials in AI/ML, (2) a clear thesis on what's becoming possible *now* that wasn't 12 months ago, (3) early signals of customer pull. Demos and decks aren't enough. We want to see real users or real research output.

Common founder mistakes

Raising too much, too early. A $5M pre-seed at a $25M valuation means your seed has to clear $50–80M valuation. Most teams aren't ready for that in 12 months. Raise what you need for milestones, not what you can get.

What we'd advise: target a 12–18 month runway, pick 2–3 measurable milestones (real users, real revenue, or real research breakthroughs), and price your round at what you can comfortably grow into.