Why 5.7 Million Americans a Year Are Filing Paperwork to Become Their Own Boss – and Why 2022–2032 Is Going to Look, in the Rearview Mirror, Like the Steamboat, the Railroad, the Model T, and 1995 All at Once

Every generation gets one of these. A technology arrives, the cost of trying something falls through the floor, and a few million ordinary people simultaneously decide they’d rather own a lottery ticket than rent one. We are in the middle of one right now, and the numbers are not subtle.

I. THE NUMBER THAT SHOULD BE ON THE FRONT PAGE

1.In 2025, Americans filed 5,671,836 applications to start a business – an all-time record, beating 2023 (5.48 million) and 2021 (5.41 million). Source: U.S. Census Bureau.

2.From 2005 through 2019, that number sat in a sleepy band of 2.4 to 2.6 million a year. We doubled it in 2020 and – this is the part everyone gets wrong – never came back down.

3.June 2026 alone: 531,423 applications, seasonally adjusted. About 17,700 a day. Roughly every five seconds, somebody in America decides they’re a company now. Filings are running 16.9% ahead of last year.

4.It is not a tech story. June 2026 by sector: retail 97,096; professional services 82,990; construction 48,933; other services 44,314; transportation and warehousing 34,512; health care 33,436; food service 28,777. Landscapers, lawyers, truckers, home health aides, taco trucks. Everybody.

5.About 30% are “high-propensity” filings – roughly 1.7 million a year showing real payroll indicators. Throw out every tax wrapper and fishing-boat LLC and you still have over a million serious swings per year.

II. THE PINK SLIP IS THE STARTING GUN

6.In 2025, U.S. employers announced 1,206,374 job cuts, up 58% from 761,358 in 2024 – the highest since 2020 and seventh-highest since Challenger, Gray & Christmas began counting in 1989. Q4 was the worst fourth quarter since 2008.

7.That same year, announced hiring plans totaled 507,647 – the lowest since 2010. Read those two lines back to back and you have the entire mood of the American workplace in one breath.

8.First half of 2026: tech alone announced 139,156 cuts, up 83% year over year, nearly a third of all U.S. job cuts.

9.AI is now cited openly. Blamed for 54,836 cuts in all of 2025, it accounted for 87,714 through May 2026 alone – about 22% of every layoff in the country, and was the single most-cited reason in March.

10.The mechanism is arithmetic, not vibes: when hiring is slow and firing is common, the risk premium on a “safe job” collapses – at the exact moment the cost of the alternative also collapsed. Safe option riskier, risky option cheaper, people move. This is an old American groove: the 1930s, the Rust Belt, 2001, 2009. Roughly 69% of employers turned to freelancers after the 2023–24 tech layoffs, so the demand side is playing along.

The layoff and the launch are the same event, viewed from opposite ends.

III. EVERYBODY IS A VENDOR NOW

11.MBO Partners counts 72.9 million Americans doing independent work in 2025 – close to 45% of the labor force, including 27.6 million full-time independents.

12.Upwork puts it at 39% of U.S. workers freelancing in some capacity in 2026, up 4 points in one year. McKinsey says 36% of employed Americans are independent, up from 27% in 2016. Three methodologies, three numbers, one unmistakable direction.

13.5.6 million independents cleared $100,000 in 2025 – up 19% year over year and nearly double the 3 million who did it in 2020. Skilled U.S. freelancers earned an estimated $1.5 trillion in 2024. This is not all subsistence gig work.

14.The structural fact underneath everything: Census data shows nonemployer firms grew 2.7% annually from 2012 to 2023 while employer firms grew 1.1%. Employers’ share of all U.S. businesses fell from 24.6% to 21.6%. Four out of five American businesses now have no employees at all.

15.The upshot: the LLC is the new résumé. The distance between “I got laid off” and “I’m consulting” is a filing fee, a Stripe account, and about forty-five minutes.

IV. IT COSTS LESS TO START A COMPANY THAN TO REBUILD A TRANSMISSION

16.State LLC filing fees run $40 to $500. That is the entire legal barrier to entry in most of the country.

17.Service businesses routinely launch for under $10,000, often under $5,000. Storefronts average near $100,000 in year one; mobile businesses about $92,500. Which tells you precisely why the growth is happening where the rent isn’t.

18.Inventory and raw materials eat 31.6% of first-year spending. Sell your judgment instead of an object and you have deleted a third of the budget before you start. Meanwhile only 30% of new owners qualify for a traditional loan and 39% fund it from personal savings – nobody is waiting for a banker’s permission anymore.

19.What used to require capital now requires a subscription. Storefront: Shopify. Payments: Stripe. Payroll: Gusto. Legal drafting, design, and analysis: an AI subscription cheaper than a phone plan. A 1908 automaker needed a factory and a foundry before selling one car. A 2026 founder needs a laptop and nerve. That is the biggest single difference between this boom and every boom before it.

V. AMERICA IS A NATION OF DEGENERATE GAMBLERS, AND THAT IS THE ENGINE

20.U.S. commercial gaming revenue hit $78.72 billion in 2025, up 9.2% – the sixth straight record – throwing off $18.09 billion in taxes. All 38 commercial markets grew.

21.Sports betting handle: $166.94 billion wagered, up 11%. iGaming added $10.74 billion, up 27.6%. And the frontier keeps moving – in May 2026 the prediction market Kalshi alone handled nearly $15 billion in sports volume in a single month, outside state regulation entirely.

22.Sit with that. Americans put $167 billion on the outcomes of games they cannot influence, coach, or play in. They are not risk-averse. They never were. The Puritans ran lotteries.

23.So here is the reframe, and I’d argue it’s the most useful idea in this article: starting a business is the only bet in America where you pick the team, coach the team, play in the game, and write off the equipment. Same neurochemistry. Far better odds. Actual tax treatment. The catch is that it’s still a bet, placed with rent money – which brings us to the section nobody puts on the podcast.

VI. WE HAVE BEEN HERE BEFORE – FOUR TIMES, TO BE EXACT

24.STEAMBOATS (1820s). Gibbons v. Ogden (1824) broke the Fulton–Livingston monopoly on New York waters. Operators flooded in, fares cracked, river freight rates fell hard. New technology plus a broken gatekeeper equals a stampede. The pattern is two centuries old.

25.RAILROADS (1830–1916). The U.S. had 23 miles of track in 1830. 2,808 by 1840. 30,000 by 1860. 163,597 by 1890. It peaked at 254,037 miles in 1916. Between 1868 and 1873 alone the country laid 29,589 miles. Hundreds of railroads were chartered; most were absorbed or went broke. The map of America was redrawn by what were, technically, startups.

26.AUTOMOBILES (1900–1929). There were 253 active American automakers in 1908. By 1929 there were 44. Meanwhile registered cars went from 8,000 in 1900 to 468,000 in 1910 to 23 million in 1929. Note the shape: a massacre of firms alongside a nearly 3,000-fold explosion in the actual thing.

27.THE INTERNET (1995–2001). Same shape, faster clock. Pets.com died. Amazon didn’t. The fiber, the protocols, and the habits all survived the bust regardless of who owned them.

28.The pattern is identical every time: a general-purpose technology drops the cost of doing something → entry explodes → most entrants die → survivors become enormous → the country ends up dramatically richer than before. The mania is not a bug in the process. The mania is the process.

29.2022–2032 is that script with AI in the lead – except entry costs less than in any prior cycle. No shipyard, no right-of-way, no foundry, no server rack. The 253 carmakers of 1908 each needed a building. The 5.7 million applicants of 2025 needed a Wi-Fi password.

VII. THE PART NOBODY PUTS ON THE PODCAST

30.BLS survival data, barely changed in decades: 20.4% of new businesses fail in year one. 49.4% are gone by year five. 65.3% by year ten. The information sector has the worst first-year odds; agriculture the best, at 6.9% first-year failure and 50.5% ten-year survival. Choose the battlefield before the battle.

31.The good news in the same dataset: the hazard drops fast once you clear year one – roughly 21% fail in year one, then only 11% in year two, 8% in year three, 6% in year four. Survival compounds.

32.Now the part that should genuinely worry you. Capital is not democratizing; it is concentrating harder than at any point in venture history. U.S. venture funding hit $412.7 billion in the first half of 2026, with 86 cents of every dollar going to AI. Rounds of $100 million and up made up 87.5% of all capital deployed, versus 33.1% in 2025.

33.OpenAI and Anthropic alone absorbed roughly 43% of global startup funding in H1 2026. Deal counts are flat or falling. Fewer companies, bigger checks.

34.So there are now two separate startup economies: a dozen capital-drenched AI giants, and 5.7 million people bootstrapping on a credit card. The middle is vanishing. If your plan involves a conventional Series A for a non-AI business, you need a different plan – not a better deck.

Plan to be profitable, not funded. That is the whole 2026 lesson in six words.

VIII. WHAT 2032 ACTUALLY LOOKS LIKE

35.Start with the official forecast, the most under-discussed number here. BLS projects the economy adds just 5.2 million jobs from 2024 to 2034 – total growth of 3.1%, versus 13.0% the prior decade. Self-employment is officially projected to grow a whopping 2.2%.

36.Read that against 5.7 million applications a year. The government’s own forecast says the jobs aren’t coming, and millions of Americans have already priced that in. I’d bet against the 2.2%, and it isn’t close.

37.Applications settle into a 5.5–7 million band through 2032. The pre-2020 baseline is gone permanently; it was an artifact of friction, and the friction is gone.

38.The median new firm gets smaller and richer – a large class of one- and two-person companies doing $200,000 to $2 million in revenue, with AI doing what used to take three hires. The solo eight-figure business stops being a novelty story by 2030.

39.Law and benefits chase the money. Portable benefits, worker classification, and state pilots become a real national political fight by 2028. When 45% of the labor force is independent, the New Deal employment model is a rounding error from irrelevance.

40.The 2026–2028 AI shakeout mints second-time founders – the most dangerous people in any economy, because they arrive with scars, a customer list, and no illusions.

41.Boring wins. Health care, home services, skilled trades, and logistics will quietly generate more durable new-business wealth than software. Construction filed 48,933 applications in a single month. Nobody writes think pieces about drywall contractors. They should.

42.By 2032, a large minority answer “what do you do?” with a portfolio instead of a title – two clients, an equity stake, and a thing they sell on the side. That is not a gig economy. That is a merchant class, and America hasn’t had a real one since the corporation swallowed everybody around 1950.

IX. IF YOU’RE GOING TO DO IT ANYWAY – AND YOU ARE

43.Start it while you still have a paycheck. The most boring advice in business remains undefeated. Overlap, don’t leap.

44.Sell it before you build it. Applications are free. Customers are not. Those 5.7 million filings are the entry fee, not the game.

45.Pick a field where the 65% ten-year failure rate doesn’t apply to you, because you have something unfair: a license, a rolodex, a reputation, or a skill people already pay you for. Nobody beats the base rate with enthusiasm. You beat it with an edge.

46.Keep fixed costs near zero for twelve months, and size it like the bet it is. Businesses rarely die of bad ideas; they die of rent and payroll arriving before revenue. You wouldn’t put the mortgage on a six-leg parlay – but unlike the parlay, this one you get to influence every single day.

THE BOTTOM LINE

In 1908 there were 253 American automobile companies, and any sensible person could see most were doomed. Those sensible people were exactly right – 209 of them were gone within twenty years. They were also, in hindsight, watching the single best decade in American history to go build something with wheels on it.

We are living in the 1908 of something. The entry fee is a laptop and a filing fee. The failure rate is roughly what it has always been, which is to say survivable. And the difference between the people who look back on 2022–2032 with regret and the ones who look back grinning comes down to something remarkably unglamorous:

Who actually filed the paperwork.

Sources: U.S. Census Bureau Business Formation Statistics (June 2026 release); Challenger, Gray & Christmas; BLS Business Employment Dynamics and Employment Projections 2024–2034; MBO Partners State of Independence 2025; Upwork Future Workforce Index; American Gaming Association; PitchBook-NVCA Venture Monitor; Crunchbase.