Raising capital for an AI company
Investors in an AI raise check three things beyond the usual: that you have the right to the data you train on, that the model and its outputs are yours to sell, and that margins hold after compute costs. Put data licenses, model lineage, compute contracts and evaluation results in clearly labeled folders.
Folder index
- 01 Overview Deck, product, where the company sits in the model stack.
- 02 Data rights Data sources, licenses, consents, customer data use terms.
- 03 Models Model lineage, third-party and open source models used, their licenses.
- 04 Evaluation Benchmarks and internal evaluations with methodology.
- 05 Compute Cloud and GPU contracts, commitments, credits and expiry dates.
- 06 Unit economics Inference cost per unit, gross margin by product.
- 07 Customers Contracts, usage data, AI-specific terms and indemnities.
How this raise works
AI companies range from model builders to application companies built on third-party models. Investors include generalist and AI-focused venture funds, strategic investors such as cloud and chip companies, and growth funds. Rounds can be large relative to revenue because training and compute are expensive.
Data and model rights are the core legal diligence. Investors ask where training data came from, under what license or consent, whether customer data is used for training and on what terms, and which open source or third-party models the product depends on and under which licenses. Unclear rights can block a round or a later sale.
Economics need their own proof. Inference costs move gross margin, so investors want cost per request or per customer, compute commitments and credits with their expiry dates, and how pricing covers usage. Evaluation results, and how they were measured, help investors judge claims about model quality. Companies selling into the EU also field questions on the EU AI Act.
Best data rooms for this raise
The 16 rooms re-ranked with this raise's weights. Investor tracking carries the most weight here.
- Investor tracking 25%
- Permission control 23%
- Raise workflow 21%
- Investor onboarding 16%
- Cost of a raise 16%
-
Full deal room with AI, a listed price and a 14-day free trial.
Price $149 a month Trial 14-day free trial Features 8 of 12 listed -
Long-running M&A platform with layered permissions and fast support.
Price Priced on request Trial Free trial available Features 9 of 12 listed -
Deal room with AI and readiness tools for preparing a process.
Price Priced on request Trial Free trial available Features 8 of 12 listed -
Bank-grade platform for large processes, with AI and a mobile app.
Price Priced on request Trial Demo on request Features 10 of 12 listed -
Simple flat-rate deal room with signing and Q&A.
Price $250 a month Trial Free trial available Features 8 of 12 listed -
One of the oldest deal rooms, common on bank-led transactions.
Price Priced on request Trial Demo on request Features 8 of 12 listed -
European deal room strong in real estate, with AI features.
Price Priced on request Trial Demo on request Features 8 of 12 listed -
Easy, reliable deal room favored by mid-market advisers.
Price Priced on request Trial Free trial available Features 7 of 12 listed -
Deal room with a listed monthly price and an open API.
Price $299 a month Trial Free trial available Features 7 of 12 listed -
Content cloud with watermarks and signing, priced per user.
Price $15 a user a month Trial Free trial available Features 9 of 12 listed
Places 11 to 16: SmartRoom (7.0), DealRoom (7.0), ShareVault (6.9), Venue by DFIN (6.3), Digify (6.2), DocSend (5.9)
Folder index for the room
A starting index for this raise. Number the folders so the order survives exports and investors can cite them in questions.
- 01 OverviewDeck, product, where the company sits in the model stack.
- 02 Data rightsData sources, licenses, consents, customer data use terms.
- 03 ModelsModel lineage, third-party and open source models used, their licenses.
- 04 EvaluationBenchmarks and internal evaluations with methodology.
- 05 ComputeCloud and GPU contracts, commitments, credits and expiry dates.
- 06 Unit economicsInference cost per unit, gross margin by product.
- 07 CustomersContracts, usage data, AI-specific terms and indemnities.
- 08 Safety and policyAcceptable use policy, safety testing, regulatory review.
- 09 SecuritySecurity reports, model and data access controls.
- 10 Corporate and financialsCharter, cap table, statements, model.
Features this raise needs
| Feature | Why | Ellty | iDeals | Ansarada | Datasite | SecureDocs | Intralinks |
|---|---|---|---|---|---|---|---|
| View-only files | Files can be read but not saved or printed, and the setting can change mid-raise without a new invite. | ||||||
| Investor activity log | A per-investor record of opens, pages and minutes, which tells the IR team who is in diligence and who stalled. | ||||||
| Q&A module | LPs and lead investors post questions on a file; the deal team routes each one, drafts a reply and decides who sees it. | ||||||
| Dynamic watermarks | Stamps the reader and a timestamp over every page, so a PPM that leaks carries its source with it. | ||||||
| AI features | Help the vendor lists, for example summaries, search across documents, automatic filing or draft answers. |
Setting up the room
- 01
Trace every training source
List every training data source with its license or consent basis before the raise.
- 02
List model dependencies
Name the third-party and open source models used and attach their license terms.
- 03
Show margin after compute
Present gross margin with inference and hosting costs per product.
- 04
Explain evaluations
Share benchmark results with the method, so investors can judge them.
- 05
Protect weights and code
Never put model weights in the room; offer supervised technical reviews instead.
Common mistakes
- No record of the license basis for training data.
- Compute credits counted as permanent cost savings.
- Customer contracts that allow training on customer data without clear terms.
Calculators for this raise
Or take the four-question quiz and check the raise checklist.
Questions and answers
What do investors look for when an AI company raises?
Data and model rights, defensibility, gross margin after compute, customer traction, evaluation results and the team.
Do I need to share model weights?
No. Offer architecture documents, evaluation results and a supervised technical session instead.
How should compute credits be shown?
Separately, with amounts and expiry dates, and with a margin view that excludes them.