An earn-out is the clause where both sides agree to keep disagreeing.

On January 12 the Delaware Supreme Court, in Johnson & Johnson v. Fortis Advisors, told J&J that it owed the former stockholders of Auris Health hundreds of millions of dollars for milestones that were never reached, under a contract providing that the money was due only if they were reached. J&J had paid $3.4 billion for Auris in 2019 and promised up to $2.35 billion more if the company's surgical robots cleared ten hurdles by fixed dates. None was cleared. The Court of Chancery had nonetheless entered a judgment of $1,011,271,291 in October 2024, and the Supreme Court removed one slice and affirmed the rest. The agreement contained no promise to hit anything. It contained a promise to try, and a definition of trying. Why would anyone sign that?

Buyers and sellers sign it constantly. An earn-out is the clause two parties write when they cannot agree on what a company is worth and would rather close than keep arguing. The buyer pays part of the price now and promises the rest if the business hits a number later. The seller, who believes the number, takes the promise. The buyer, who does not, keeps the cash. Each leaves the room convinced it has won. One annual study found earn-outs in 22 percent of 2024 private-target deals outside life sciences, and its next edition found 24 percent for 2025.

After closing, one side holds the pen. On October 26, RadNet filed with the SEC the merger agreement to buy OnRad, a teleradiology company, through a subsidiary, for $350,000,000 in cash, before adjustments, plus five million RadNet shares, according to the 8-K. Section 2.7(b) says: "The Earn-Out Amount shall equal the product of (i) fifteen (15) multiplied by (ii) the lesser of (A) the amount (if any) by which the Post-Closing Annual Adjusted EBITDA exceeds $56,000,000 and (B) $14,000,000." Every dollar of adjusted earnings between $56 million and $70 million is worth fifteen dollars to the sellers, up to $210 million, and a dollar below $56 million is worth nothing. The buyer prepares the statement. The sellers' representative gets thirty days to object and thirty more to negotiate, after which an accounting firm named in the agreement, RSM US LLP, decides the disputed items in a decision the contract calls "conclusive, binding, and non-appealable."

At fifteen times, a $100,000 disagreement about how to allocate shared overhead is a $1.5 million disagreement about price. The number is always wrong, and here every error counts fifteen times.

The agreement also says how OnRad will be run. The buyer promised to run the company as "a separate business unit," with its own books, managed day to day by a Key Employee the agreement identifies only on a schedule the filing leaves out, under the general oversight of the buyer's chief executive. It also promised not to "take any action, or omit to take any action, the primary purpose or primary intent of which is to circumvent, reduce, delay, or avoid" the payment. The definitions of the earnings measure and of the earn-out period sit in Annex C-1, which the filed copy references but does not include; the 8-K says the year ends December 31, 2027. The clause nobody read is here a clause nobody outside the deal can read in full.

That "primary purpose" language is the market. The 2025 study found that among 2024 deals with earn-outs outside life sciences, 90 percent carried a covenant that the buyer would take no action whose primary purpose was to defeat the payment, 10 percent carried a commercially reasonable efforts standard, 5 percent a promise to maximize the payment, and 3 percent a promise to run the business as the seller had. The 2026 edition puts the 2025 figures at 82 percent and 16 percent. Delaware's high court read a clause of that kind in 2015, in Lazard Technology Partners v. Qinetiq North America Operations. QinetiQ paid $40 million for Cyveillance and promised up to $40 million more on a revenue target. Section 5.4 barred the buyer from taking "any action to divert or defer [revenue] with the intent of reducing or limiting the Earn-Out Payment." Revenue fell short. The sellers argued that the buyer knew its decisions would sink the earn-out, and the Supreme Court held that the clause "only barred the buyer from taking action specifically motivated by a desire to avoid the earn-out." Knowing was no breach.

The sellers had asked for more. Their drafts would have obliged the buyer to maintain existing levels of business, fund the company adequately, recruit enough staff, bid for new contracts consistent with past practice, and divert nothing. The buyer struck every one, and the court declined to restore them under the implied covenant of good faith, because the seller "had sought objective standards for limiting the buyer's conduct but lost at the bargaining table." So the clause in most deals is of a kind the seller has already lost under before it signs.

Auris got what Cyveillance's sellers were refused. J&J agreed to use commercially reasonable efforts to achieve each regulatory milestone, and the agreement defined the term by J&J's own habits, as efforts "consistent with the usual practice of Parent and its Affiliates with respect to priority medical device products of similar commercial potential at a similar stage in product lifecycle." A separate clause barred decisions "based on taking into account the cost of making any Earnout Payment(s)." An inward-facing standard like that usually favors the buyer, until the buyer is the biggest in its industry and the yardstick is its own priority programs. The Court of Chancery, in Fortis Advisors v. Johnson & Johnson, found one comparable priority device, an orthopedic robot called Velys, and found that iPlatform "received starkly different treatment than Velys." Within days of closing, Auris's iPlatform robot was in a contest J&J called Project Manhattan, after which the winner was merged with the loser and, in the trial court's phrase, "effectively became a parts shop for Verb." A priority device, the court held, "would not have to endure a costly battle merely to remain operative," and the Supreme Court agreed.

By my arithmetic the reversal removes $300 million of the $900 million in contract damages, the first milestone's $400 million weighted by the 75 percent chance of approval the parties had assigned it. That leaves $600 million on the contract and, with the $60.9 million fraud award, about $661 million before interest. The case went back to Chancery to recalculate the interest, and on January 26 the court entered a stipulated final judgment of about $811 million, Fortis's lawyers announced. J&J's report for the quarter ended September 27 says the company paid a $0.8 billion judgment, inclusive of interest, that January. I do not know what OnRad will earn in 2027, and neither does anyone who signed on October 26.

Say Crowfield Industrial Holdings buys Pemberton Gasket for $80 million plus three times any EBITDA above $10 million in the first year. In month two Crowfield moves Pemberton onto its group accounting system and charges it a shared-services fee. In month five it signs a group supply contract that lowers Pemberton's prices to a large customer in exchange for volume at a sister plant. Each is ordinary, each lowers Pemberton's EBITDA, and neither has the primary purpose of avoiding the earn-out, because each has an ordinary purpose of its own. After the year ends the sellers receive a statement showing EBITDA of $9.6 million, and they have thirty days.

Primary purpose is a fact about the inside of a buyer's head. The person who has to prove it no longer has a badge to the building.

Both sides agree because an earn-out is the only clause that lets each side be right about the price, and the disagreement they could not settle is moved to a date when one of them owns the business, keeps the books, and writes the first draft of the number. The other holds a thirty-day objection window and a word, and the word is intent. The earn-out prices the seller's willingness to argue about that word later, from outside the building.

Columns are the views of their authors. They are not legal advice and not the position of the firm.