Minerva, the Y Combinator-backed accounting startup, posted on its own account that it bought a single accounting practice and took its operating margin from 5% to 70% by rebuilding it AI-native. Nobody has audited that number independently. It is still the number that should worry an accounting firm owner more than any acquisition offer.

Here is why. Private equity has been buying accounting and law firms since 2021, hundreds of them, and it has sold almost none. Nikola Lazarov, co-founder and CEO of Eilla AI, laid out the scoreboard in his September 23 newsletter, Services as Software, and it is close to empty. A typical UK buyout holds a position around six years, which means the 2021 and 2022 vintage, the first wave of top-20 accounting and law firm deals, should be coming to market right now. It mostly is not.

The buyer with the checkbook changed

The 2021 buyer had a simple thesis. Add partners, bill more hours, bolt on smaller firms, exit to a bigger fund in five or six years. Fifty more partners in two years was good news under that thesis, because more partners meant more billable hours to sell.

The buyer with capital today is not running that thesis. Current, backed by Thrive Holdings, committed more than $500 million to buy accounting firms and rebuild their tax work on OpenAI’s models. Minerva says in that same post that its bought firm now runs on a fraction of the headcount it started with.

This buyer, in Lazarov’s framing, splits the firm into three pieces. The client relationship keeps its value. So does whoever holds the license, since in most US states an audit practice legally has to be owned by a licensed CPA and a law firm by a licensed attorney. Everything in between, the associates, seniors and managers, is exactly what a 2021 fund called growth. This buyer calls it cost.

We already covered each half of this separately. In Thrive committed a billion dollars to accounting firms, AI dissolves the billable-hour asset the 2021 thesis paid for. In how the MSO structure buys the layer regulation left unprotected, ownership law draws the line between the licensed professional and everything else, and that is where value moved. This piece is what happens when two buyers price both sides of that line, and the prices never touch.

One flip. Two funds buying themselves out

Exactly one change of ownership between funds is confirmed. In January 2025, Blackstone bought New Mountain Capital’s stake in Citrin Cooperman, the position New Mountain took in 2021. Terms were not disclosed. Figures Lazarov reports, unconfirmed by either party, put New Mountain’s 2021 entry at roughly 11x EBITDA on about $315 million in revenue, and Blackstone’s price at close to 15x with revenue nearing $850 million, a valuation around $2 billion. Four times the money in three years, if those numbers hold up.

Everything else labeled an exit is a different transaction wearing the same word.

EisnerAmper closed a continuation vehicle with TowerBrook in March 2026, its investor since 2021. TowerBrook did not sell to an outside buyer. It raised a new fund to buy its own position and cash out whichever of its own investors wanted out. EisnerAmper is now the 13th-largest US firm, at $1.2 billion in revenue and 27 acquisitions since 2021. It grew. It did not change hands.

Fletchers Solicitors did the same with Sun Capital in February 2026, that fund’s first continuation vehicle. Fletchers grew EBITDA from £8.0 million in 2021 to £37.9 million by September 2025, adding ten acquisitions along the way. Another firm that multiplied in size, still sitting in the same portfolio.

The one real sale to an unrelated buyer I could find is smaller. Investcorp bought Stowe Family Law from Livingbridge in September 2024 for roughly £48 million, a specialist family-law practice, not one of the top-20 firms the 2021 wave targeted.

One flip. Two funds paying themselves. One smaller genuine sale. That is the entire scoreboard.

Growth was the pitch. Now it’s the liability

Run this from the seller’s side.

A 2021-vintage owner trying to exit in 2026 walks in with a growth story. Partner headcount that multiplied, bolt-on acquisitions, more people billing more hours. That story was built for a buyer who pays by the hour.

The buyer holding capital today is the other one. And to that buyer, the headcount the seller grew is precisely the line item it plans to compress the moment the deal closes. It is not buying the firm the seller spent five years building. It is buying the client list and the license underneath it, and pricing the rest as overhead.

That mismatch, more than any lack of buyer appetite, is why the scoreboard is this empty. Someone wants to buy. They just do not want to pay for what a 2021 seller spent five years building. The metric that decides price stopped being revenue per head. It is revenue per licensed professional now.

What changes if you run a services firm outside the US

Your firm is probably not getting a call from Blackstone next year. The same arithmetic is reaching you anyway, just from your clients instead of from a buyer.

If you still measure your firm’s growth by how many people you added to bill more hours, you are building exactly the layer an AI-native competitor can price below you tomorrow. That competitor does not need fifty more consultants. It needs the client relationship and whoever signs the work with a professional license, and it runs execution on models underneath.

The concrete number to pull this week is revenue per person who legally has to sign the work (a licensed accountant, a bar-admitted attorney, a certified engineer), not revenue per employee. If the second number is climbing faster than the first, your firm is fattening the layer that gets commoditized first instead of the one that holds your price.

You do not have to guess which layer is which. It shows up in the operating data your firm already has.

Let’s map where your firm’s pricing power actually lives