The Antitrust Division That Stopped Going to Court

On 7 July, Bloomberg published a tidy piece of arithmetic about America's mergermergerWhen two companies combine into one, which can reduce competition if the two were rivals. police. Since January 2025, it reported, the second Trump administration's Antitrust DivisionAntitrust DivisionThe part of the US Justice Department that enforces competition law, including deciding whether to challenge mergers in court. has litigated to block a merger exactly once, and that one case settled before trial. Over the same stretch it reached 'more than a dozen' settlements. The Federal Trade Commission, the other US merger cop, sued three times (winning one case, losing another, with a third headed to trial). The story's frame, taken from the official now running the division, Associate Attorney GeneralAssociate Attorney GeneralThe number-three official at the US Justice Department; in 2026 this official, Stanley Woodward, took day-to-day charge of the Antitrust Division. Stanley Woodward, is that merger review had become 'a tax on dealmaking.' Woodward, the Justice Department's number-three official, took day-to-day charge of the division after Gail Slater, the Senate-confirmed and more enforcement-minded antitrust chief, was pushed out in February 2026.

That framing invites an intuitive conclusion, one you may already have drawn: a Justice Department that stopped suing must be letting deals through on the cheap, trading the hard remedy for the soft one. In antitrust the hard remedy is structural, forcing the merged company to sell off a real, standalone business so a competitor survives. The soft remedy is behavioral, also called conduct: a promise to play nice, such as licensing a technology, walling off data, or pledging not to discriminate against rivals. Structural remedies cut; behavioral remedies supervise. Enforcers and economists have long been skeptical of the second kind, because a promise is only as good as the years of monitoring behind it.

So the natural story writes itself: fewer lawsuits, more settlements, and the settlements are hollow. We went to check whether that natural story is true. It is not, and the way it fails is the more interesting story.

Two numbers in that report look like they contradict each other: sued once, but settled six times. They do not contradict, and the reconciliation is the whole point. Under the Tunney ActTunney ActA 1974 US law requiring the government to publicly file and justify any antitrust merger settlement so a judge and the public can scrutinize it., the government files a complaint alongside almost every merger settlement. So a settled case still technically involves a 'complaint.' The meaningful distinction is between a contested, litigated challenge, where the government sues to block and actually squares up for trial, and a complaint filed the same day as a pre-negotiated consent decreeconsent decreeA court-approved settlement in which the merging companies agree to conditions instead of fighting the government at trial., where the fix is agreed before anyone lawyers up. By the first, courtroom meaning, the division went to trial footing once. By the second, it papered six settlements. Both are true; they count different things.

One caveat keeps that 'once' honest. A second of the six decrees, UnitedHealth's purchase of the home-health company Amedisys, also began as a lawsuit to block. But that suit was filed in November 2024, under the previous administration, and this division inherited it and settled it with divestitures in August 2025. So the count of genuinely new blocking suits the second-term division chose to bring is one: HPE/Juniper.

One more gap to close. Bloomberg counts 'more than a dozen' settlements; we code six. The difference is scope. Our six are the formal merger consent decrees filed under the Tunney Act, the population we can read and label with precision. The larger, looser 'dozen-plus' folds in matters we cannot code the same way: non-merger conduct matters, deals restructured or abandoned under pressure without a formal decree, and informal resolutions. We coded the set that self-labels. Where we cannot see the paperwork, we do not guess.

Of the formal merger settlements the Antitrust Division actually signed, how many are real divestitures and how many are just promises?

This question has a clean, public answer, because the US government is required to show its work. Every time the division settles a merger case, it must file the settlement in federal court under the Tunney Act and publish a Competitive Impact StatementCompetitive Impact StatementThe document the government must file under the Tunney Act explaining exactly what a settlement requires and why it protects competition., a document that says, in the division's own words, exactly what the merging companies must do and why it protects competition. These filings are self-labeling. You do not have to guess whether a remedy is a divestituredivestitureSelling off a business unit to a buyer, the core of a structural remedy. or a pledge; the paperwork tells you.

So we did the boring, decisive thing. We pulled every Tunney Act merger settlement the division filed from January 2025 through this week, read each one, and coded it into three buckets: structural (a business gets sold to a named buyer), mixed (a divestiture with a behavioral condition bolted on), or behavioral-only (nothing gets sold; only promises are made).

Before looking, we committed to a prediction that could be proven wrong: coded from their own filings, structural divestitures would outnumber behavioral-only decrees by at least two to one. If behavioral-only settlements turned out to be the majority, the hypothesis wins and our claim dies.

It did not die. Five clean divestitures. One mixed case. Not a single behavioral-only decree.

A caution against over-reading our own result: six is a small number, and a two-to-one bar is easy to clear at that size. We are not claiming statistical significance. What we are claiming is stronger and more literal, because it is a census, not a sample: this is the complete set of formal merger decrees the division has signed, and in that complete set, behavioral-only appears zero times. You cannot argue with a category that is empty.

Coding is only honest if you can see the codes. Here is every settlement, with the business that changed hands, in plain language.

The lone mixed case, HPE/Juniper, is the one the wire counts as the single litigated suit: the division sued to block it on 30 January 2025, then settled on 27 June 2025, roughly two weeks before the scheduled 9 July trial. Even here the structural spine is real, HPE had to divest an entire global 'Instant On' Wi-Fi business, but a behavioral condition rides alongside it: the parties must auction off a license to the source code of Juniper's 'Mist AI' network software, and if the bidding clears $8 million with multiple bids, Juniper has to license at least one more buyer. A pure divestiture sells a business and walks away; this decree also polices HPE's future conduct. That is why we code it mixed rather than structural, and it is the closest thing in the whole set to a behavioral remedy, bolted onto a divestiture rather than standing alone.

If the remedies did not weaken, the deal-friendly posture that everyone senses has to live somewhere. It does, just not where the 'hollow promises' story looks. The softening is upstream of the remedy, in the prior question of whether a deal gets challenged at all.

Look at what cleared with nothing attached. Paramount Skydance's roughly $111 billion combination with Warner Bros. Discovery, and Nexstar's roughly $6.2 billion purchase of Tegna, both closed with no conditions, and in both cases reporting says senior officials waved them through over the objections of the career staffcareer staffThe non-political, long-tenured lawyers and economists at the agencies who investigate deals, as distinct from the political appointees who make final calls. who were investigating them. In the Paramount matter, Variety reported that department leadership cleared the deal before the line lawyers, who were leaning toward suing, could formally register their concerns. That is the softening: not a weaker remedy, but no remedy, because there was no challenge.

And the count of serious, contested challenges is falling. Dechert's DAMITTDAMITTThe Dechert Antitrust Merger Investigation Timing Tracker, a law-firm dataset that measures how long significant merger investigations take and counts significant litigated challenges. tracker, which measures how long significant merger investigations run and counts the litigated ones, shows US investigations stretching to record length even as the number that end in a courtroom fight shrinks. Fewer fights, longer waits, and when a fight does end in a decree, it still ends in a divestiture.

The honest sentence is not 'divestitures gave way to promises.' It is 'divestitures or nothing, and more often nothing.'

One last honesty check, turned on our own finding. Coding a decree 'structural' says what the government required on paper. It does not prove competition was actually preserved. A divestiture only works if the business that was sold survives and competes under its new owner, and that verdict does not arrive on signing day. It arrives years later, when you can see whether Viavi kept Spirent's testing lines competitive, whether BrightSpring and Pennant ran the divested home-health sites as real rivals, whether the buyer of HPE's Wi-Fi business made anything of it.

The mixed case is already contested. A coalition of 13 state attorneys general, led by California, moved to intervene against the HPE/Juniper settlement in October 2025, calling it facially inadequate, and a federal judge held a rare Tunney Act hearing in March 2026 to weigh whether the decree is truly in the public interest. So even the one settlement with a behavioral component is not settled in the deeper sense; its adequacy is being litigated by the states the federal division did not fully satisfy.