Updated
For funds and quant firms
Sell Hedge Fund Data to AI Companies
Short answer
Yes, if the firm owns what it licenses. Some AI data programs ask for multi-year proprietary time series or trading infrastructure code, plus the right to license it for AI training. Exchange feeds you only redistribute and public data don't fit. Have your lawyer check your data licenses before anything goes in.
To sell hedge fund data to AI companies, you license proprietary data or code that your firm owns. A license gives a program permission to use a copy. You keep the originals and keep ownership.
The hard question for a fund is ownership. A long price history may rest on an exchange feed you only rent. This page lists what programs want, which firms fall short and which licenses can stop you.
DillisData is a data agency that matches firms with AI data programs. We show why each match fits. The fund owes DillisData nothing, because programs pay us a referral fee.
- Only what you own
- Live signals can stay out
- No charge from us
What programs ask for
What do AI companies look for in hedge funds?
Rules vary by program and can change, so treat this list as a guide to what some request. A time series is a set of values recorded in order over time. Trading infrastructure code is the software that sends orders, reads market messages and tests ideas.
| What some programs ask for | Why they want it |
|---|---|
| Multi-year proprietary time series that your firm owns. These can cover market, risk or sports data. | Value depends on history. A private series that runs for years is hard to find anywhere else. |
| Or trading infrastructure code that your firm owns | Firms keep these systems private, so they are hard to find in public code. The code itself can be the asset. |
| The right to license it for AI training | This is the deciding test. Holding a copy of data does not mean you may license it. |
| A firm of about 5 to 500 people, in programs built around this kind of data | These programs look at what the firm built more than how many people it has. |
| Roughly 15+ people in office roles, for whole-firm programs. Desk jobs are the only ones counted. | These programs study a team at work, so they need enough people on desks together. |
| For those same programs, 5+ tools in steady use, like chat, email, files and a CRM, with 3+ years of history | Linked tools show how decisions got made around the work. Offers fall when tools are left out. |
Who is not a fit
Which funds and quant firms are not a fit?
Better to find this out now than after a long legal review. Some programs turn away the items below.
Redistributed exchange feeds you don't own
Exchange data you pay to receive is usually not yours to pass on. Storing years of it does not change that. A tick archive built only from a vendor feed is unlikely to pass.
Data that is already public
Programs can get public prices, filings and news on their own. A cleaned copy of public data is not what they ask for here.
Records your firm doesn't own
Clean ownership matters. Code or data from a former partner, a vendor or an outside manager falls short until the rights are clear.
Defense and export-controlled work
Some programs won't accept this at all. Most funds won't hold it, but check any portfolio or research files that touch those fields.
Your records
Which fund records can add value, and where do they live?
Some programs say these can raise an offer. The names below are standards and tools found in many trading shops. Your stack may differ. If you hold mostly code, our page on licensing source code goes deeper.
| Records | Where they often live | What they show |
|---|---|---|
| Order management systems | Private code repositories | How orders get checked, routed and tracked |
| Handlers for FIX, ITCH and OUCH | Private repos. FIX is an open trading standard. ITCH and OUCH are Nasdaq protocols. | How software reads market messages and enters orders |
| Low-latency engines and backtesters | Private repos and test servers | How a firm builds for speed and tests ideas on past data |
| Research notebooks and backtest results | Jupyter notebooks and shared drives | How a researcher moved from an idea to a result, failures included |
| Proprietary time series | Databases such as kdb+ | Values your firm built or recorded, tracked over years |
| Sports odds and outcome histories | In-house databases | How prices moved and what really happened. Odds bought from a vendor carry that vendor's terms. |
Rules to clear
Can exchange data licenses stop you from licensing?
Yes, they can, and they are only one of the checks. This is general information, not legal advice. Bring this list to your lawyer and your chief compliance officer.
Exchange data and AI training
At least one exchange now speaks to AI directly. Nasdaq's AI policy says any use of its data, including training AI models, must follow your license. It says trained models may be fee-liable and may need added licenses. Other exchanges and vendors set their own terms.
Series you built from licensed data
A signal built on exchange prices may still be tied to that license. Your own math does not always clear the input. Ask your lawyer to review each series against each feed contract.
Who owns the code
Copyright Office Circular 30 addresses this. Work an employee creates in regular duties is generally a work made for hire. The employer then counts as the author. Contractors are different, and it can take a written agreement. Check every outside developer and any quant who brought code from a past job.
Secrets inside the code
Repos collect broker logins, API keys and connection strings. GitHub secret scanning is one tool that looks through commit history. Treat any key it finds as burned, and replace it.
Inside information
Research notes can hold material nonpublic information, such as facts from a company or an expert call. Section 204A of the Advisers Act says covered advisers must keep written policies to prevent its misuse. Have your compliance officer screen those notes.
Investors and their records
Investor names, statements and account details belong on your exclusion list. If you are SEC-registered, Regulation S-P may cover personal facts about individual investors. Fund papers and side letters can add secrecy terms too.
Your edge
This one is a business call. Once a model has trained on your data, you can't get it back. You may want to keep live signals and current positions out. You could limit the license to older or retired work.
Get a figure before the legal review starts
A short quiz covers your staff, years of records, systems and region. A number shows on screen. You name no strategies and hand over no code.
The process
How does a fund go from first look to a signed license?
Our part is the match. We never hold your data or code. You sign with the program directly, and any transfer is between you and that program.
Tap through the quiz
It takes 8 to 14 single-tap answers for most people. Nothing asks about your positions or your models.
2 minutesReview your estimate and top 3
One estimate appears, along with your 3 closest program fits and the reasons for each. We point out terms that need a close read.
Sort what you own
List each series and repo. Mark what you built, what rests on a feed and what you would never share. Your lawyer tests that list against your contracts.
Sign and hand over
Your pick may offer one-click sign-up, or its representative may reach out. It then copies what the contract allows, strips personal details and pays the firm.
Public facts
What public filings and reports show
These are public market facts, and DillisData was part of none of them. None of them is about a fund, and none predicts a price.
Google's bid when it won an auction for old Spirit Airlines emails, chats and files. The sale still awaited court approval in early November 2026.
Axios, Aug 2026Reddit's data licensing deals from January 2024, which run two to three years, per its IPO filing.
SEC Form S-1, Feb 2024Publisher Wiley's total from AI deals so far, per its June 2026 filing. This money comes in bumps from one quarter to the next.
Wiley 8-K via SEC EDGAR, Jun 2026Timing
Should a fund wait, or look at this now?
A look costs nothing, and a decision can wait for your lawyer. These points are why a look is worth doing now. How long current prices last is unknown.
Private series are the scarce part
Data no program can find elsewhere earns the most. Once a competitor supplies something close, demand for the next set tends to fall.
Retired work earns nothing
A strategy you closed years ago still sits in storage. Its code and backtests may be worth more licensed than left alone.
Exchange terms keep changing
Nasdaq has already written AI into its data policy, and others may follow. Knowing today what your contracts allow helps you avoid a surprise later.
Program prices can shift
Pay can shift as more firms join. Labs also test other training methods, and the date of any change is unknown.
What might your private data and code bring in?
Programs typically pay a fund $100K to $5M+, largely depending on fund size. See that as a typical range, never a promise. Your estimate gives you a figure to weigh before you spend on lawyers.
Questions
Questions owners ask
Can a hedge fund license its tick data?
Only if the fund owns it or holds the right to license it. Ticks taken from an exchange feed usually come with limits on passing them on. Programs turn down redistributed feeds a firm doesn't own. Ticks your own system generated, such as your own order records, are a separate question for your lawyer.
Can we license code without any data?
Some programs say yes. They ask for multi-year proprietary time series or trading infrastructure code, so either can qualify. Order management systems, FIX, ITCH and OUCH handlers, low-latency engines and backtesters can all add value. The firm still has to own the code.
How much history do we need?
Some programs ask for multi-year series, and value depends on history. They don't give one fixed number for all data. Whole-firm programs often ask for 3+ years of records. The quiz checks your years against each program's rule.
Could a license reveal our strategies?
It could, and you should weigh that first. You choose the scope, and the contract records what stays out. Training can't be undone. A careful start is retired strategies and older research, with live work kept out.
Do small quant shops qualify?
They can. Some programs built around this data look at firms of about 5 to 500 people. Programs that want whole-firm records may ask for about 15+ office staff instead. A small shop with strong owned code may match the first kind only.
Does sports data count?
It can. Some programs value sports odds and outcome histories that a firm owns. Odds you bought from a data vendor come with that vendor's license, so check it first. Public scores alone won't meet the test, since public data doesn't fit.
Do our investors need to know?
That depends on your fund papers and side letters, so ask your lawyer. Investor personal details should stay out either way. You may choose to tell them even if no contract requires it.
Does DillisData see our code or data?
No. We ask general questions about the firm and test the answers against program rules. Your code and data never reach us. If you sign, they go straight to the program. For the bigger picture, read how businesses sell data to AI companies.
Next step
See what your owned data could earn
Answer the quiz, get your number and take it to your lawyer. You name no strategies. If the figure doesn't move you, walk away.
No DillisData charge. No code shared. Not an offer.