Compute: the defining cost
Training and experimentation can consume a large share of a round. Investors will review how compute estimates were built and whether they include failed runs. Research credits, academic compute access and partnerships can lower costs, and some companies negotiate committed-use discounts once usage is predictable.
University IP and spin-out terms
If your technology came from a university, it may own the IP and license it to the company in exchange for equity, royalties or both. These terms affect how much of the company is left for founders and investors. Negotiate them early and get them in writing; investors will want to see the final agreement.
- University equity stake and whether it dilutes in later rounds.
- Royalties or revenue shares on products.
- Field-of-use limits on what the licence covers.
Deep-tech investor terms
Investors backing technical risk may ask for milestone tranches, technical advisers on the board or information rights on research progress. These can be reasonable, but make sure milestones are ones you can reach with the money provided.
Talent and equity
Research engineers are expensive and in demand. Expect a larger option pool than a typical software company, and account for its dilution before and after the round.
Costs and terms: applied AI versus ML-first
| Item | Applied AI startup | ML-first company |
|---|---|---|
| Biggest cost | Sales and product | Compute and research talent |
| IP risk | Customer and data contracts | University or prior-employer ownership |
| Typical extra terms | Standard | Milestone tranches, technical board roles |
| Option pool | Standard | Often larger for research hires |
Questions to prepare before you pitch
- Is the university stake subject to dilution in future rounds?
- What milestones would trigger each tranche?
- What information rights on research progress do you need?
- How large an option pool do you expect, and pre- or post-money?
How KJ Enterprises evaluates machine learning businesses
KJ Enterprises considers ML teams with clean IP, realistic compute budgets and milestones tied to a commercial product.
If that describes your company, you can apply for investment. Related reading: the complete funding guide and the step-by-step how-to and applied AI costs and terms.