Score your assetRegisterAsk

Run an AI company with real contracts and real hardware? Here is what an investor can actually buy.

If you have enterprise contracts, racks you own, or both, you have assets that behave like assets everywhere else. Weights and datasets do not. This page is about financing the first kind without selling the story.

$305M+
in assets structured and supported
100+
companies, on four continents
20+
jurisdictions we have worked in
2018
founded, eight years doing only this
Artificial intelligence
What AI company tokenization changes for you.

Your problems, and what we do about them

Why AI company tokenization is slow today, and what fixes each problem.

Written in the owner’s words. If yours is not on the list, the readiness score will find it in about eight minutes, and nobody calls you.

Your problemWhat we doWhat you get
01

Investors cannot tell the model from the company

The record shows the company: cap table, contracts, hardware, with the evidence attached.

A buyer who sees what is actually for sale.

02

Compute is a photo, not an asset register

Equipment title, invoices and location enter the record at their tier.

Hardware a lender can lend against.

03

Contracted revenue is scattered across inboxes

Customer contracts with payment terms sit next to the revenue they produce.

One schedule an investor can read in an afternoon.

One asset, three jobs

The record, the raise and the token, one after another.

The record

Intelligence

Because an investor needs to see the company, not the demo.

Equity reconciled to the paper

Articles, register, option ledger and every note, with the disagreements between them named. The AI part changes none of this. The speed of your last round is what made it messy.

Compute treated as the equipment it is

Title to the machines, the invoices, where they physically sit and the terms of the room they sit in, each with its document. An investor reads what you own, where it is, and what it is worth on paper.

Revenue with the contracts behind it

Contracted revenue taken from the accounting system, with each customer agreement attached, so the figure and its source can be read together.

How Intelligence works →

The raise

Raisable

A raise against what customers have already signed

The memorandum and the risk factors are drafted from the contract schedule in the record, so an investor reads terms rather than projections.

Equity or a share of revenue, decided by your lawyer

Whether you sell part of the company or a share of contracted revenue is a structuring decision your counsel makes. The package is prepared to whichever standard applies.

A flat fee for the window, never a percentage

A licensed broker-dealer runs any regulated sale. No success fee, no carry, no cut of what you raise at any layer of ours.

Example. Say you want to fund the next two racks without giving away more of the company. Your lawyer structures a share of contracted revenue. The package describes the contracts and the hardware from the record, eligible investors verify and sign on your own address, and the register records each commitment as it lands.

How Raisable works →

Tokenization

Compass

Because a payment rule the contract runs beats a report somebody remembers to send.

Payments run from the register

A share of revenue pays everyone holding on the day the rule names, in USDC, a digital dollar, and each payment is written down as it happens. Nobody runs a report first.

What you getA share of revenue paid by a rule, not by somebody running a report.

If it is equity, the register is the list of owners

Current at the moment of every transfer, with the shareholder agreement enforced by the contract instead of remembered.

What you getA share register that is the asset itself, if what you sell is equity.

Who may hold it is a condition of the asset

Written into the token and checked on transfer. There is a path to secondary trading through licensed venues, and listing rests with the venue.

What you getA check on who may buy, run by the asset on every transfer.

How Compass works →

Side by side

AI company tokenization side by side: the same four questions, before and after.

Nothing here needs the rest of the page. If the right column is not worth the work to you, the answer is no, and that is a fine answer.

How it works todayHow it works after

What is being sold

A story about the model

Equity, machines or contracted revenue, with evidence

How investors are paid

A report, then a transfer

From the register, by the rule, recorded

What a buyer checks

The deck, and your word

Contracts, invoices and titles in the record

Who may hold it

Whoever you vet by hand

Whoever the rules admit, checked as it moves

Our clients

AI company tokenization case studies: named work, and the marks that go with it.

We have not yet taken an artificial intelligence company through a tokenized raise. What we have done is finance the physical side of compute, three times, and that is the honest comparison.

Egypt · Fundraising strategy

Pylon

We structured a securities offering that financed hardware for a Y Combinator-backed company, using its contracts rather than another equity round.

Equipment with contracted revenue behind it can be financed on its own terms.

Germany · Token structuring

Decentraliced

We built a financing structure that balanced the yield from computing hardware, the cost of its own power supply, and what investors were owed.

Compute and the electricity it burns are one calculation, not two.

United States · Fundraising strategy

Quarry Dynamics

We structured a raise that paid for computing equipment and the site it runs on, in Canada.

Hardware, site and power can be raised for together, as one project.

What you will be asked for

Documents needed for AI company tokenization, and who has to have issued them.

Every answer in the record carries its document and a proof tier from T0 to T5. A number you typed and a number from a registry both get in, and they are not stored as though they were equal.

Certificate of incorporation and articles

T5

A registry

Share register and option ledger

T3

Your books

Title to the hardware, invoices and location

T4

Signed by the other side

Customer contracts with payment terms

T4

Signed by the other side

Revenue from the accounting system

T3

Your books

Where this usually stops: stage 05 · Issue. The existing cap table has to be reconciled with reality before it becomes chain state. The six stages are on the Compass page. T5 is issued by an authority, T4 is signed by the other side, T3 is your accounting system of record; anything asserted without a document sits at T0, and the score will say so.

Updated 3 October 2026

Talk to us

Is this right for your AI company?

Tell us the asset and the jurisdiction. A specialist answers in plain words, and says what is still in development.

In plain English

AI company tokenization terms: six words this page uses, in plain English.

You do not need any of this to talk to us. It is here because these words get used at you in meetings, and knowing them is worth more than nodding along.

Compute

The machines that run the models: graphics cards, servers and the building and power they need. Unlike a model, it is a physical thing you can own and finance.

Contracted revenue

Money customers have already signed up to pay you over a period, as opposed to money you expect. Lenders and investors treat the two completely differently.

Depreciation

How quickly the hardware loses its value. On compute it is fast, and any honest financing structure has to say what happens when it does.

Cap table

The list of everyone who owns a piece of the company. It is the asset in a private company, and an AI company’s cap table has usually been through several fast rounds.

Exemption

The rule that lets you sell to investors without a public registration. Your lawyer picks it, and it decides who is allowed to take part.

Model weights

The trained output of the model itself. They are hard to own cleanly, hard to transfer and hard to value on any given day, which is why we do not build offerings on them.

When this is not for you

When AI company tokenization is not for you: three cases, in advance.

Said here so you can leave without a call. It is cheaper for both sides than discovering it in week six.

×

A company before revenue with no hardware of its own. That is an ordinary venture round.

×

Model weights, datasets or intellectual property you do not clearly own, as the thing being sold.

×

A network token or a community allocation. It is a security or it is nothing.

Primary sources: Securities Act, section 4 - exempted transactions – 15 U.S.C. 77d · Regulation D, Rule 506 – 17 CFR 230.506.

Three steps to a call

Find out where it stands.

01

Score your asset

Twenty-five questions across seven dimensions, about eight minutes, no email to see the result.

Take the Readiness Score
02

Ask the founder, live

The monthly webinar with Gene Deyev: 40 minutes on Zoom, once a month.

Join the monthly webinar
03

Book a call

Bring the asset and the score. We will say what is missing, and say so if the answer is no.

Book a call