Nobody Closes a Brokerage. They Just Stop Paying for the Bond.

Over the past two years, 10,003 freight brokerages lost their operating authority. We pulled the federal paper trail on every single one of them — the authority events, the insurance filings behind the authority events, and the state incorporation records behind those — to answer a simple question: how does a brokerage actually die?

The answer is more uniform than any industry death we’ve ever measured: 99.8% of them died exactly the same way. Involuntary revocation — which for a broker means one thing. The $75,000 bond lapsed, and FMCSA pulled the authority.

Not bankruptcy filings. Not voluntary wind-downs. In two years of data, the number of brokers who formally surrendered an active authority is one. Everyone else just stopped paying, and the system did the rest — fast. Matching each revocation to the specific bond filing that died first, the median gap between the bond’s cancellation date and the revocation is 8 days. And 97% of those dying bonds were flat cancellations, never replaced. When a broker’s bond goes, they don’t shop for a new surety. It’s over.

The body count, by the numbers

Trailing 12 months ending	Exits	New entrants	Net
July 2024	9,322	5,843	−3,479
July 2025	6,172	4,711	−1,461
July 2026	4,001	3,932	−69

Active broker authorities peaked at 30,936 in October 2022. Today the count sits around 24,700 — a 20% contraction. The single worst month was October 2023: 1,058 brokerages revoked in 31 days.

But look at that bottom row. The purge is over. The last twelve months are essentially flat, and several recent months were actually net positive — the first growth since the peak. Hold that thought, because the flat line is not what it looks like. We’ll come back to it.

They died at 25 months old

The median brokerage that died in this window was 25 months old — one bond-renewal cycle plus a freight recession. Forty-four percent never reached their second birthday.

By the year they got their authority:

Authority granted	Share revoked in the last 2 years
before 2010	2.1%
2019	7.3%
2021	8.9%
2022	13.8%
2023	30.5%
2024	53.3%

Read that last row again. More than half of every broker authority granted in 2024 was already revoked within two years. The 2021 boom class is now as dead — 83% gone — as broker cohorts from before 2010, which had two full decades to accumulate their mortality. We showed last May that the COVID cohort was dying at four times the base rate. This is what the mechanism looked like from the inside.

And the deaths weren’t just young — they were born to die. Among dead brokers we could link to state incorporation records, the median company got its federal authority 3–4 months after the LLC was formed. These weren’t established businesses adding a brokerage arm. They were formed to be brokers, during the two years when anyone with a laptop could be one.

One more incorporation-record detail worth a long stare: brokers registered through United States Corporation Agents — LegalZoom’s budget registered-agent service — are 46.9% dead today. Brokers registered through CT Corporation, the blue-chip corporate agent, are 6.7% dead. A seven-fold survival gap, visible from nothing but how the company was formed.

Where the bodies are

Death rates against the brokers active in August 2024 (30.1% of them exited overall):

Geography	Exit rate
Georgia (worst state)	38.5%
Texas	35.9%
Florida	35.4%
South (region)	34.4%
Midwest (region)	24.5%
Wisconsin	19.0%
Minnesota	18.3%
Nebraska (safest state)	12.0%

Fresno brokerages died at 44%. Las Vegas at 42%, San Antonio at 41%, Atlanta at 40%. Laredo and Houston posted the biggest absolute body counts. And the pattern holds within states, not just across them: metro-ZIP brokers died at 35.6% versus 25.2% for rural ones.

Put the vintage table and the geography table together and the die-off stops looking like uniform industry attrition. It’s one specific animal dying: the 2021–2024 metro Sun Belt startup, formed cheap, bonded cheap, dead at two years old. A three-fold survival difference separates a Nebraska brokerage from a Georgia one. The old rural Midwest shops barely noticed the apocalypse.

The resurrection door — and the dog that didn’t bark

Eighteen percent of the recent dead had died before. FMCSA’s revoke-and-reinstate door swings freely: 1,840 of the window’s bond-losers had lost their bond at least once previously, and the all-time champion — a Georgia outfit called Rapid Carmovers — has lost its bond eight times since 2012. The five-or-more club even includes dockets belonging to Maersk’s Performance Team unit and DSV, where repeat revocation on a barely-used docket reads as billion-dollar administrative sloppiness rather than distress. The same table holds both the world’s biggest logistics companies and a guy in Loganville, and that’s the point: the bond is the only wall, and it’s a low one in both directions.

Here’s what we expected to find and didn’t. We’ve documented how carriers reincarnate — die under one authority, re-emerge under a fresh one with a clean record. Brokers don’t. Of 10,003 dead broker authorities, exactly 26 — 0.3% — re-entered under a new MC number on the same USDOT. A broker’s docket carries no safety record worth escaping, so the dead either resurrect the same docket or stay dead.

Mostly they stay dead — and they don’t even clean up. Among dead brokers linkable to state registries, 80% of the legal entities were never dissolved. Still “Active” with the Secretary of State, bond unpaid, authority revoked, owner gone. Ten thousand walk-aways, and almost no shared officers or addresses connecting them. This wasn’t fraud rings. It was ten thousand individually rational decisions to stop paying $75K worth of premium for a business that no longer cleared it.

Now, about that flat line

Since early 2025 the broker count has been stable. It would be easy — and the trade press has mostly found it easy — to read that as recovery.

Here’s what the stable market actually contains. While ten thousand small brokerages were dying of bond lapse, the largest brokers in America spent the same recession rebuilding themselves around AI, on the record and in numbers:

• C.H. Robinson cut headcount roughly 30% from its 2022 peak — while its truckload volume grew, while it logged eleven consecutive quarters of market-share growth, and while shipments per person per day rose more than 50%. Quote turnaround went from 17–20 minutes of human work to about 31 seconds, answering 100% of quote requests instead of the 60–65% a human desk got to. Their CFO, on an earnings call: “if today we’re doing 600,000 requests for freight quotes and it goes to 6,000,000, we don’t have to add anyone back.”
• RXO posts 15–19% annual gains in loads per person, with AI agents making half a million phone calls a quarter.
• Industry-wide, brokerage-sector employment has been dead flat since 2024 while 3PL net revenue grew 5% — the same freight, moved by fewer people.
• Digitally-native brokerages run an estimated 40–50% lower carrier-ops cost per load than manual desks.

Two honest caveats, because our numbers only stay credible if we flag what they can’t say. First: those productivity figures are company-reported and unaudited, and cutting staff mechanically flatters any per-person metric. Second: nothing in the data shows AI caused the broker count to stabilize — the purge ended because the vulnerable cohort was already dead, and every analyst postmortem we reviewed attributes the shakeout to plain economics: rates, overcapacity, credit.

But here’s the asymmetry that should keep a small broker up at night. The giants have published, quarterly-repeated proof that they can grow volume while shedding a third of their people. For the 24,000 small brokerages that survived the purge, there is no data at all — the flagship industry surveys didn’t even ask an AI question until this year, and not one of the AI-tooling vendors selling into small brokerages has published a verifiable customer metric. The only brokers with receipts are the ones taking your freight.

The last extinction event had a mechanism: a bond payment that stopped making sense at recession margins. The next one has a candidate mechanism too: a quote that costs your competitor 31 seconds and costs you 20 minutes. The 2023–24 filter killed brokers who couldn’t cover their overhead. Nothing in this data says the next filter waits for a recession to start running.

What to take from this

1. A brokerage death is a bond death. If you want to see the die-off coming — for a counterparty, a customer, a competitor — watch the insurance filings, not the revenue rumors. The paper trail runs bond-cancel → revocation in eight days, and one revocation proceeding in three gets cured at the brink. The bond filing history is the whole early-warning system.
2. Don’t read the flat count as a healed market. Half the brokers registered in February 2022 are gone. The stable-looking total is churn at steady state — smaller entering classes replacing a still-dying tail — inside a market whose biggest players just rebuilt their cost structure.
3. Survival correlated with everything except being in a freight hub. Old beat young, rural beat metro, Midwest beat Sun Belt, CT Corporation beat LegalZoom. Every marker of the pre-boom, capitalized, relationship-anchored brokerage beat every marker of the boom-era startup. The market didn’t punish brokers. It punished a vintage.

---

Receipts: Authority counts are active broker operating authorities, reconstructed by replaying every grant, reinstatement, revocation, and disposition event in FMCSA’s authority history file (data complete through May 2026 — the federal feed lags about two months). Exit = an active authority interval ending; 2-year window = Aug 2024–Jul 2026. Bond forensics join each exit to FMCSA L&I insurance history (BMC-84/85 filings). Incorporation data from bulk Secretary-of-State records (~20% linkage; link rates near-identical for dead and surviving brokers, so comparisons are fair). C.H. Robinson and RXO figures from Q4 2025 / Q1 2026 earnings transcripts and 10-K filings; cost-per-load estimate from FreightWaves (Jan 2026). Methodology and queries available on request.