On September 11, 2026, the Census Bureau published a number almost nobody wrote about. Americans filed 531,728 business applications in August, down 7.8% from July. The release ran four pages, got no press conference, and disappeared into the economic calendar. It is also the freshest hard evidence available on whether 2026 is a good time to start a business.
A different number has been everywhere since spring. Between November 2025 and January 2026, U.S. entrepreneurs filed 1.56 million business applications, the largest three-month stretch since the Census Bureau started tracking the series in 2004. That figure has anchored roughly every “now is the moment” article published this year.
Both numbers are real. Both are being read wrong. A business application is a request for an Employer Identification Number filed with the IRS on Form SS-4, not proof that a company exists. So the 1.56 million does not count businesses, and the August dip does not signal a collapse. The gap between paperwork and a functioning company is the single most useful thing a prospective founder can understand about this market.
Is 2026 a good time to start a business? Yes, for a specific kind of business, and the Census Bureau’s own projections explain exactly which kind.
Last updated: September 2026
Quick answers
Is 2026 a good time to start a business?
Yes, for businesses that can reach revenue without payroll or outside funding. Census Bureau data shows filing volume at structural highs, AI tooling costs at record lows, and small business credit meaningfully tighter than 2019. Lean service and software models fit those conditions. Inventory-heavy and headcount-heavy models do not.
How many new businesses started in 2026?
Roughly 3.66 million business applications were filed in the first eight months of 2026, up about 17.8% year over year. Applications are not businesses. The Census Bureau projects that only about 5% of any given month’s filings become employer businesses within four quarters.
Is it harder to get funding in 2026?
Yes. Full small business loan approval rates have fallen 21 percentage points since 2019. Large banks approve roughly 13% to 15% of small business loan applications, and SBA 7(a) loans carry interest rates of about 10% to 15% with a practical credit floor near 680.
Is 2026 a good time to start a business?
2026 is a good time to start a business if your model reaches paying customers without hiring employees or borrowing money, and a poor time if it requires either. That is the honest version of the answer, and it comes directly out of two datasets moving in opposite directions.
On the cost side, everything got cheaper. The Small Business & Entrepreneurship Council’s March 2026 survey found 82% of small business employers have already put money into AI tools, running a median of five tools across their operations, with 93% planning to spend more. Work that used to require a hire now requires a subscription.
On the capital side, everything got harder. Full approval rates for small business loans sit 21 percentage points below where they were in 2019. Large banks approve somewhere between 13% and 15% of applications they receive. The practical credit score floor for an SBA loan has moved to roughly 680, and 7(a) rates run 10% to 15%.
Put those together and the market is telling you something specific. It is not saying start a business. It is saying start a business that does not need a bank.
Founders who internalize that constraint early tend to build differently from the ones who spend six months writing a plan that assumes a credit line. If you are still deciding what to build, GJ’s roundup of 101 business ideas for 2026 is organized around exactly this filter.
What the record filing numbers actually count
Filings and companies are tracked as two separate series, and the distinction is the Census Bureau’s own, not an interpretation. The agency’s Business Formation Statistics report applications monthly and employer formations separately, because an EIN request tells you someone filed a form and nothing more about whether they ever opened for business.
The August 2026 release makes the gap concrete. Of 531,728 total applications, the Census Bureau classified 145,387 as high-propensity, meaning they carry characteristics historically associated with becoming a payroll business. Only 34,263 listed a planned wages date. And the projection for how many of August’s filings will become employer businesses within four quarters came in at 28,501.
Run the division. About 27% of August’s filings were high-propensity, 6% planned to pay wages, and 5.4% are projected to make payroll inside a year.
The gap is enormous.
That does not mean 94% of filers fail. Most new businesses in America are non-employer businesses: freelancers, consultants, solo operators, single-member LLCs formed for a side project. They never appear in the employer formation series because they never hire anyone. They can still be profitable, and increasingly they are the point rather than a consolation prize.
So when an article cites 1.56 million filings as proof that entrepreneurship is surging, the accurate translation is that 1.56 million people requested a tax ID. A large share of them are building something that will never have a second employee. That is the actual shape of the 2026 boom, and it matches what solo founders building million-dollar AI businesses have been demonstrating for two years.
How many new businesses started in 2026?
Americans filed about 3.66 million business applications through the first eight months of 2026, roughly 17.8% ahead of the same period in 2025. Monthly volume has moved between roughly 496,000 and 580,000 all year, against a 2019 monthly average near 290,000.
The month-to-month numbers are noisy and that noise gets over-interpreted in both directions. January 2026 came in at 532,319. February dropped 5.8% to 496,443, which produced a small round of “the boom is over” commentary. May recovered to roughly 524,000. August landed at 531,728 after a 7.8% decline from July.
Strip the volatility and one fact holds: filing activity has settled at close to double the pre-pandemic rate and has stayed there for four consecutive years. The high-propensity series tells the same story. Americans filed about 1.72 million high-propensity applications in the twelve months through June 2026, roughly 31% above the 2019 level.
Industry detail is where the August release gets interesting. Retail trade applications fell 26.0% month over month to 104,087, the sharpest drop in the report. Construction slipped 1.4% to 48,120 and professional services were essentially flat at 80,691, down 0.4%. Manufacturing, wholesale trade, transportation, information, and arts and entertainment all rose.
Read that industry table as a map of where capital intensity hurts. Retail requires inventory, leases, and working capital, the three things a 13% bank approval rate punishes hardest. Professional services requires a laptop.
Why it got cheaper to start and harder to fund
The cost of building software and running a business’s back office collapsed in 2026 while the cost of capital did not. By early September 2026, the Silicon Data LLM Token Expenditure Index hit a record low near $0.97 per million tokens, down from roughly $2.04 at the end of May. Andreessen Horowitz has tracked this curve for two years and calls it LLMflation, estimating inference costs fall on the order of 10x annually.
Turbo AI is the clearest illustration of what that enables. Rudy Arora and Sarthak Dhawan, two twenty-year-old college dropouts, launched an AI note-taking tool in early 2024 and grew it to five million users by October 2025, with eight-figure annual recurring revenue, profitability, and a team of about 15 people in Los Angeles, according to TechCrunch. Twenty thousand new users join daily. Their user base includes analysts at Goldman Sachs and McKinsey.
That company would have needed a far larger team and a funding round in 2019. In 2026 it needed neither.
The same force is pushing people toward the exits at large employers. Challenger, Gray & Christmas found that AI was cited as the primary reason for more than 49,000 planned job cuts through April 2026, about 16% of all announced cuts, up from 5% across all of 2025. Block cut 4,000 jobs in February, dropping from more than 10,000 employees to under 6,000, and Jack Dorsey told shareholders that “a significantly smaller team, using the tools we’re building, can do more and do it better.”
Not everyone buys the framing. Critics pointed out that Block tripled headcount between December 2019 and December 2022, and that unwinding a hiring binge is not the same as being replaced by software. Both things can be true. For a prospective founder the practical read is identical either way: large companies are getting smaller, and the tools making that possible are available to you at retail prices.
Is it a bad time to start a business right now?
It is a bad time to start a business that depends on cheap credit, a long runway before revenue, or a team of employees from day one. Those three dependencies are all more expensive in 2026 than at any point in the last decade, and the credit market is not hiding it.
The borrowing picture is unambiguous. Large banks approve roughly 13% to 15% of small business loan applications. Community banks land at 19% to 22%, credit unions at 20% to 24%, and CDFIs and microlenders reach about 72%, though at much smaller loan sizes. SBA 7(a) rates run 10% to 15%, and lenders effectively require a 680 credit score after the 2026 SBSS cutoff increase.
The broader credit environment is under strain too. Roughly $586 billion in debt comes due for S&P 1500 companies during 2026, and U.S. corporate bankruptcies have reached a 14-year high. For an incumbent carrying debt, that is a threat. For a new entrant with no debt, it is the competition thinning out.
Failure rates deserve the same honesty. Bureau of Labor Statistics data shows 20.4% of new U.S. businesses close within the first year, 49.4% within five years, and 65.3% within ten. LendingTree’s analysis of the same data puts first-year closures at 22.1%. The widely repeated “90% of startups fail” claim does not appear in any federal dataset.
Roughly four in five new businesses survive their first year. That is a better base rate than most people assume, and it has not meaningfully changed in 2026. What changed is which business models can reach year two without external money. Anyone weighing the personal side of this tradeoff should read GJ’s breakdown of the real pros and cons of being an entrepreneur before filing anything.
Which business models fit 2026 conditions
The models that fit 2026 share one trait: the first dollar of revenue arrives before the first dollar of fixed cost. Everything in the Census and lending data points at that structure, and the industry-level filing numbers show founders already sorting themselves accordingly.
| Model | Typical startup cost | Time to first dollar | Outside capital needed | 2026 fit |
|---|---|---|---|---|
| Solo services and consulting | Under $1,000 | Days to weeks | None | Strong. Professional services filings held flat in August while retail fell 26%. |
| AI-assisted micro-software | $300 to $5,000 | 1 to 6 months | None to minimal | Strong. Inference at $0.97 per million tokens makes the cost structure work solo. |
| Inventory retail and ecommerce | $10,000 to $100,000+ | 2 to 9 months | Usually required | Weak. Working capital needs collide with a 13% to 15% bank approval rate. |
| Employer-track business with payroll | $50,000+ | 3 to 12 months | Almost always | Mixed. Only 6.4% of August filings even listed a planned wages date. |
None of this makes inventory businesses unbuildable. It makes them expensive to build with borrowed money in a year when borrowed money is scarce, which is a different and more fixable problem. Presales, deposits, and consignment all sidestep it. GJ’s ranking of the top 10 businesses to start in 2026 and its list of 50 small business ideas both sort options by capital requirement, which is the variable that matters most right now.
How to tell if 2026 is your year
Three conditions separate founders who fit this market from founders who are fighting it. Each one maps to a specific number in the data above, which is what makes them testable rather than inspirational.
Condition one: you can reach your first paying customer without payroll. Only 5.4% of August 2026 filings are projected to become employer businesses within a year. That is not a failure statistic, it is a description of what this economy is actually producing. If your plan requires two hires before revenue, you are building against the grain, and the lending data says the grain will not bend for you.
Condition two: you can fund the first twelve months without a loan. With large banks approving 13% to 15% of applications and SBA 7(a) money priced at 10% to 15%, a plan that assumes debt financing is a plan with an 85% chance of needing a new plan. Founders who start inside a job, on savings, or on customer deposits control their own timeline.
Condition three: AI removes a real cost from your model, not just a task from your day. The 82% AI adoption rate among small business employers means the tools are table stakes, not an edge. The edge belongs to founders who can point at a specific line item that would have been a salary and now is not. Turbo AI reached five million users with about 15 people. That ratio is the bar.
Hit all three and the timing question answers itself. Hit two and you have a real business that needs a different funding structure. Hit one and the honest move is to keep working on the model rather than the filing.
The record everyone cites is not why 2026 works. The cost structure is. A generation of founders is discovering that the cheapest company they will ever build is the one they can start this year, and Census filings show 43% of Gen Z already acting on it, a pattern GJ tracked in its report on Gen Z entrepreneurs in 2026. The paperwork is easy. Deciding what the paperwork is for is the part worth spending a weekend on, and GJ’s guide to starting your first business in a single weekend is built for exactly that.
