What is an applied-AI product studio worth?

An applied-AI product studio sits in an unusual spot between an agency and a product company: it bills clients to build bespoke AI-powered software, but unlike a pure development shop, it sometimes keeps a piece of what it builds — retained equity in a client’s product, a licence to reuse part of the underlying platform, or usage-based royalties on something it shipped years ago. That blend is exactly what makes it hard to value with a single multiple. A studio that is really just billing hours dressed up in AI language is worth roughly what any agency is worth; a studio that has genuinely retained ownership in things it built is worth something closer to a small, illiquid portfolio company, and the two can look identical on a one-page summary of last year’s revenue.

What a buyer is actually paying for

The most valuable asset a studio can show is a portfolio of shipped products that are still generating usage or royalty revenue well after the delivery engagement ended, because that is proof the studio’s work outlives the invoice. Retained IP or equity positions in client products add real value, but only to the extent they are actually documented and transferable — a verbal understanding that the studio 'kept a piece' of a client’s company is not an asset a buyer can price. A reusable internal AI stack — evaluation harnesses, prompt libraries, fine-tuning pipelines built once and reused on every new engagement — is a genuine efficiency asset that shortens delivery time on the next project, and named technical founders whose reputation drives inbound project leads are themselves part of what is being bought, for better or worse.

Why work-for-hire agreements can leave almost nothing behind

A studio that signs standard work-for-hire agreements assigning all IP to the client on every engagement is, in substance, a development agency that happens to use AI tools — there is no retained product asset behind the revenue, only the pipeline of future client work, which is worth considerably less than a portfolio of owned or co-owned products. This is the single most common gap between how studio founders describe their business and how it actually prices: 'we build AI products' sounds like product ownership, but if every contract assigns the output fully to the client, the studio owns nothing from any past engagement except the case study.

Why speculative equity is worth less than it looks on the balance sheet

Studios that take equity in early client ventures instead of, or alongside, cash fees often carry those stakes on an internal summary at an optimistic valuation with no real market to test it against. Illiquid stakes in ventures that may never have a liquidity event are a genuinely different kind of asset from cash-generating retained IP, and a buyer will discount them heavily, or discount them to close to nothing, unless there is a credible path to the studio actually realizing cash from the position. The presence of equity stakes on a studio’s books is not, by itself, a value driver — their quality and liquidity is what determines whether they add anything at all.

What normalizes the earnings of a studio like this one

Recasting starts with separating billable delivery revenue from any royalty or licensing income tied to retained products, since the two behave completely differently and a buyer needs to see them apart to judge how much of the business is repeatable agency work versus durable product income. One-off costs tied to a single flagship project — a large contractor engagement, an unusually expensive compute bill for one client’s model training — should come out of the base year rather than be read as a normal ongoing cost. Reliance on one or two flagship case studies for most of the studio’s inbound credibility is also worth flagging explicitly, since it is a concentration risk that affects durability even when it does not show up as a line item on the income statement.

Why two similar-looking studios price differently

Two applied-AI studios can post the same trailing revenue and still be different assets. One has three shipped products still earning usage fees, a documented internal tooling stack that halves delivery time on new engagements, and IP terms that clearly leave some ownership with the studio. The other bills the same revenue purely through work-for-hire delivery, holds no retained product, and depends on one or two named founders for every piece of inbound work. The first is closer to a small product company with services attached; the second is a services business that happens to use AI tools. Both are legitimate, sellable businesses, but pricing them the same because their revenue lines match ignores exactly the distinction that determines what a buyer is actually acquiring.