TL;DR: Citadel Securities reported 5.6 million new US business applications in 2025, up 24% since ChatGPT launched. The headline reads like a boom. The math reads like a filter. Bureau of Labor Statistics data shows roughly 80% of new businesses survive year one, 70% survive year two, and 50% survive year five. AI lowered the cost of starting a business. It did not lower the cost of running one.

The Number Behind the Headline

PYMNTS reported in July 2026 that Citadel Securities found 5.6 million new US business applications in 2025, a 24% increase since ChatGPT launched in November 2022. The research note, citing US Census Bureau data, called it evidence that AI enhances America's "hustle culture." That is one reading.

Here is another: only 30.1% of those 5.6 million applications are "high-propensity" filings, meaning the Census Bureau estimates they are likely to become actual employer businesses. In 2019, that figure was 37.6%. Christopher Olivier Consulting documented the decline. The ratio of real operating businesses to total filings is dropping, not rising.

More people are starting businesses. Fewer of those businesses are designed to employ anyone beyond the founder. That is not a boom. That is a lower barrier to entry producing higher volume at lower average quality.

The Survival Math Has Not Changed

I have watched thousands of founders pitch at Angel Investors Network since 1997. The ones who survive year two are the ones who built systems before they built products. That observation has not changed in 27 years, and the Bureau of Labor Statistics confirms it.

SCORE's compilation of BLS data shows the survival curve: roughly 80% survive year one, 70% survive year two, 50% survive year five, and 33% survive year ten. Those numbers have remained remarkably stable across decades, across economic cycles, and now across the AI adoption wave.

AI did not change those numbers. AI changed who attempts the curve. The median seed-stage startup now has 4 employees, down from 5 in Q1 2023. Companies are leaner at inception. But leaner does not mean more durable.

The 23x Gap

The data point that matters more than the 5.6 million headline is this: PYMNTS Intelligence's May 2026 Small Business Week report found that SMBs generating more than $1 million annually reported average revenue growth of 13.7%. Businesses earning less than $150,000 grew at 0.6%. That is a 23x gap. Not 2x. Twenty-three times.

Two-year survival confidence follows the same split: 93% for businesses above $1 million in revenue, 73% for businesses below $150,000. A 20-point spread in whether the founder believes their own business will exist in two years.

This is not a story about AI creating a small business boom. It is a story about AI amplifying existing advantages. The businesses that were already past the $1 million mark are compounding faster. The businesses that filed an LLC and launched a landing page are growing at less than 1% per year. AI made the starting line cheaper. It did not move the finish line.

The AI Entry Illusion

The marketing around AI-enabled entrepreneurship frames tool access as competitive advantage. It is not. Tools are infrastructure. Advantage comes from what you build on top of them.

Consider the analogy from manufacturing. The cost of a CNC machine dropped from $500,000 to $50,000 over two decades. More shops could afford one. The shops that thrived were not the ones that bought the cheapest machine. They were the ones that built production systems around it: scheduling, quality control, customer pipelines, trained operators.

AI tools follow the same pattern. A founder who connects ChatGPT to a Shopify store has the same setup as 100,000 other founders who did the same thing this month. The differentiation is not the tool. It is the operational system: customer acquisition cost below the revenue threshold, documented fulfillment processes, retention mechanics that produce repeat purchases, and financial reporting that a lender can underwrite. Those systems take 12-18 months to build regardless of what AI handles in the background.

The businesses surviving past year two in the Citadel dataset will be the ones that treated AI as infrastructure and competed on operational excellence. The businesses that treated AI as the strategy itself will be the ones filing dissolution paperwork in 2027.

Starting a Business Is Not Building an Asset

Here is the distinction that the 5.6 million number obscures. Starting a business is a transaction. Building a sellable asset is a discipline.

A founder who registers an LLC, connects an AI content tool, and launches a Shopify store in a weekend has started a business. That business has no systems, no documented processes, no recurring revenue, no team beyond the founder, and no acquirability. It is a gig with a tax ID.

Building an acquirable business requires everything the starting phase skips. SOPs that survive the founder's absence. Revenue that recurs without the founder's relationships. Financial reporting that a lender can underwrite. A team that makes decisions without asking permission. These are the assets that produce exit multiples, not the LLC filing.

The six dimensions PE buyers score on every acquisition target have not changed because AI lowered incorporation costs. If anything, the flood of new entrants makes differentiation harder, not easier. When 5.6 million businesses compete for the same customer attention, the ones with systems win. The ones with hustle burn out.

What AI Actually Changed (and What It Did Not)

AI changed three things that matter for new business formation.

First, it reduced the labor cost of early-stage operations. A founder can now handle content, customer service, and basic analytics without dedicated hires. Harvard and INSEAD found that AI-native startups employ roughly 25% fewer workers at similar valuations.

Second, it compressed the time from idea to revenue. Tools like ChatGPT, Claude, Midjourney, and ElevenLabs let founders build marketing assets, write proposals, and create product demos in hours instead of weeks.

Third, it lowered the minimum efficient scale of a viable business. Citadel Securities framed this as a structural positive. I agree with the observation and disagree with the conclusion. Lower minimum efficient scale means more businesses can start. It does not mean more businesses can sustain.

What AI did not change:

  • Customer acquisition still costs money. AI can generate content. It cannot generate demand. CAC has not dropped for most verticals.
  • Cash flow management still requires discipline. AI does not fix the founder who prices below cost because they do not understand their unit economics.
  • Operational systems still require deliberate construction. AI can help draft an SOP. It cannot build a culture that follows one.
  • Founder dependency still destroys exit value. A business that runs on the founder's daily input is worth 2-3x SDE. A business that runs on documented systems is worth 4-6x. AI did not change that math.

The Operator's Verdict

The 5.6 million number is real. The "boom" interpretation is marketing.

What the data actually shows is a market with lower barriers to entry, higher volume of participants, stable survival rates, and accelerating concentration of growth among businesses that already crossed $1 million in revenue. The businesses that are winning in 2026 are not the ones that started easiest. They are the ones that built fastest from start to system.

If you are one of the 5.6 million, the question is not whether you started a business. The question is whether you are building something that survives year two, year five, and a buyer's due diligence. AI can accelerate every step of that journey. It cannot skip any of them.

For the framework that turns a business into a sellable asset from day one, start with the Sovereignty Stack. For the 90-day audit that identifies where you are the bottleneck, read the founder dependency tax breakdown.

Doctrine Connection: Competence Beats Credentials

The doctrine applies here with precision. Competence beats credentials. Filing an LLC is a credential. Running a business that survives is competence. AI gave more people the credential. It did not give them the competence. That still requires the manual, the watchstanding, the casualty drills, and the discipline to build systems that run without you.

Frequently Asked Questions

Q: Are the 5.6 million business applications all real businesses?

No. Only about 30% are classified as "high-propensity" by the Census Bureau, meaning they are likely to become employer businesses. The rest are primarily gig registrations, side-hustle LLCs, and freelance incorporations. The total number overstates the actual formation of operating businesses.

Q: Has AI actually improved small business survival rates?

Not according to available data. BLS survival rates (80% year one, 50% year five) have remained stable across economic cycles and technology waves. AI has lowered startup costs but has not changed the underlying reasons businesses fail: cash flow problems, weak demand, and operational mismanagement.

Q: Why is there a 23x revenue growth gap between large and small SMBs?

PYMNTS Intelligence attributes it to digital fluency and multi-channel capability. Businesses above $1 million in revenue are more likely to sell across multiple channels, accept diverse payment methods, and adopt AI tools systematically. Businesses below $150,000 are more likely to operate single-channel with limited technology adoption.

Q: What should a new business founder focus on to survive year two?

Three things, in order. First, prove unit economics: ensure your product or service generates positive contribution margin on every sale. Second, build recurring revenue or repeat-purchase patterns so you are not starting from zero each month. Third, document your first three processes so the business can function for 48 hours without your direct involvement. These three steps address the top causes of year-two failure.

Q: Is the small business boom sustainable?

At the headline level, probably not at 5.6 million applications per year. At the structural level, the shift toward smaller, AI-enabled teams running profitable businesses at lower revenue thresholds is likely durable. The volume will normalize. The efficiency gains are permanent.

Jeff Barnes is the founder of Digital Evolution Marketing Group (DEMG). This article reflects operational experience, not investment advice. Results vary by business, market, and execution. Do your own due diligence.