Raising a pre-seed round in AI.
At pre-seed, AI investors back a team with an unfair insight and a workflow worth automating. The first question is always the same: why will this not be a feature of the next foundation model?
AI / ML
AI investors judge three things before anything else: an advantage the big model providers cannot copy, gross margin after compute, and proof that customers rely on the output, not just try it.
What pre-seed investors expect in AI
A demo is easy to build now. Investors look past it to three things instead.
A team that can build and ship
- Someone who has trained, tuned or deployed models in production
- Someone who knows the customer's workflow from the inside
- A working prototype built quickly and cheaply
A reason the big models will not eat you
- Access to data, users or workflows a general model does not have
- A product that gets better as customers use it
- A clear line between what you build and what you rent
A painful, specific job
- A task people do by hand today and dislike
- A buyer who owns the budget for that task
- Early users who come back after the novelty wears off
How the round comes together
Pre-seed AI money comes from people who can judge the technical bet quickly.
- Technical angels
- Pre-seed funds
- Accelerators
Angels
Often leads
Engineers and researchers from AI teams write early checks and vet the technical approach.
Pre-seed & seed funds
Often leads
Pre-seed funds with an AI thesis lead small rounds on strong technical teams.
Accelerators
Sometimes
Programs trade a small check for structure, peers and often compute or cloud credits.
University funds
Sometimes
University-linked funds back spinouts built on lab research in machine learning.
Grants & non-dilutive
Now and then
Research grants and cloud credit programs can cover early compute without dilution.
SAFE or priced round?
Almost every pre-seed AI round is a SAFE: quick to sign, valuation set by a cap. Caps in AI can move fast with the hype cycle, so know what your cap costs you at conversion before you accept a high one.
The slides investors read first
Team
Who has shipped models and who knows the customer?
Problem
The manual task and what it costs today.
Why not a model?
What a general model cannot do here, and why.
Prototype
What works now and who has used it.
The ask
Amount, instrument and the milestones it buys.
Questions founders ask
Is a wrapper on a foundation model fundable at pre-seed?
It can be, if the value sits in the workflow, the data or the distribution rather than the prompt. Investors want to hear what stays yours when the underlying model improves or its price drops.
Do I need my own model to raise pre-seed in AI?
No. Many teams start on hosted models and fine-tune or train later. Investors care more that you know when owning a model would pay off.
Sources
Sources: investors' published criteria for AI companies; cloud and model provider pricing documentation; industry reports. Editorial content, reviewed quarterly. Not financial or legal advice.
Other stages
Related sectors
Who funds this round
Learn the terms
Run the numbers
Raise in your city
Inside Unicorn Factory
Free account
Investor matching is coming to your free account. Create it now and we will match your round the day it opens. Until then, get an honest read on your deck.