The Pre-ChatGPT Stranding: How AI Funding Concentration Creates a New Underwriting Signal for Small Business Lenders

The AI boom that poured over $250 billion into OpenAI and Anthropic has produced a quiet casualty: hundreds of startups built before ChatGPT’s 2022 arrival. Those companies — many with viable products, revenue, and existing loans — now face a valuation collapse as investors pile into frontier-model builders. Lenders who extended credit to these pre-ChatGPT startups, or to the small businesses that serve them, are sitting on a portfolio blind spot.

For funders, brokers, and ISO networks, this is not a tech-sector footnote. It is a direct underwriting signal that demands action. The concentration of AI funding creates a new credit risk category – the stranded startup – that behaves nothing like a normal small business failure. Understanding the mechanics of this stranding will separate lenders who catch the exposure from those who absorb the loss.

The Mechanics of Stranding

A startup founded in 2020 with a B2B SaaS product for data labeling, customer support automation, or content generation was a credible borrower three years ago. It had a team, recurring revenue, and a growth trajectory. It may have taken an equipment loan for servers, a revenue-based loan for sales hires, or an equipment lease for office space. Its financials looked sound.

Then ChatGPT launched. The market narrative shifted. Venture capital that once spread across dozens of vertical AI applications condensed into a single bet: the foundation model. Investors told pre-ChatGPT startups that they were no longer fundable unless they pivoted toward model development or integration. Many could not pivot. Their revenue growth flattened or reversed as customers migrated to cheaper, more powerful API-based solutions.

These startups are not bankrupt. Many still generate cash flow. But they are stranded—unable to raise the next round, unable to attract acquisition interest at meaningful multiples, and often run by founders who are burning personal capital to keep the lights on. For lenders, this creates a slow-motion default risk that traditional cash-flow underwriting fails to detect.

What This Means for Funders

Direct exposure to pre-ChatGPT startups is the most obvious risk. Lenders with portfolios that include software-as-a-service loans, equipment leases, or unsecured term loans to companies founded between 2017 and 2022 must segment those borrowers by their founding date and their last institutional funding round. A business that has not raised capital since 2022 and is not generating cash flow above its debt service is a candidate for restructuring now, not six months from now.

The risk extends beyond the startup itself. Small businesses that supply goods and services to these stranded companies will feel the downstream effect. Data center contractors, cloud infrastructure resellers, marketing agencies, payroll processors, and office furniture providers all depend on a healthy ecosystem of funded startups. When a cohort of 200 stranded companies cuts headcount by 30 percent and truncates operations, the ripple hits the small business lenders who financed their supply chain.

Funders should review their concentration risk not just by industry vertical, but by “AI exposure cohort.” A borrower whose top three clients are pre-ChatGPT startups is functionally underwriting its own survival on the fate of those clients. That risk is not captured by a standard cash-flow analysis that looks backward 12 months. It requires forward-looking client dependency scoring.

Implications for Brokers and ISO Networks

Brokers who originate loans for tech-adjacent small businesses now have a new filtering criterion. When a borrower presents as a “technology services” company, ask for the founding date and the date of its last venture round. If the company was founded before December 2022 and has not raised capital since, treat it as a heightened credit risk until the borrower demonstrates a clear path to non-VC revenue growth.

For brokers serving the AI supply chain, this is also an opportunity. Stranded startups that still have revenue but lost VC support are often desperate for working capital to bridge to profitability. They may be willing to accept higher cost of capital. But the broker’s job is to ensure the borrower can actually service the debt from operating cash flow, not from a future funding raise that will never come. Underwrite to the unit economics, not the pitch deck.

ISO networks can build lead scoring models that flag borrowing companies with a “pre-ChatGPT” profile. A business with an SIC code in software publishing, data processing, or computer systems design that was founded in 2020 and has less than $2 million in annual revenue should be flagged for manual review. The network that integrates this signal into its originations platform will reduce default rates before the loan is funded.

A New Underwriting Variable: The AI Disruption Date

Standard small business underwriting relies on historical cash flow, credit score, and collateral. Those inputs are backward-looking. The AI stranding is a forward-looking event that renders past performance misleading. A company with 18 months of stable revenue can still be an imminent default if its market is being eaten by a foundation model that costs a fraction of what the company’s product charges.

Lenders should add an underwriting variable: the “disruption vulnerability score.” This score combines founding date relative to ChatGPT, reliance on proprietary data or models that are now commoditized by large language models, and customer concentration among other pre-ChatGPT startups. Tools like Crunchbase, PitchBook, and public SEC filings for the company’s investors can help generate this score.

For example, a B2B data-labeling startup founded in 2021 with $500,000 in monthly recurring revenue and a $2 million term loan from an alternative lender may appear creditworthy. But if its largest client is a pre-ChatGPT NLP platform that just lost its Series B, that revenue stream is at risk. The lender should require the borrower to disclose its top three clients and examine their funding status.

Operational Playbook for Lenders

1. Segment your existing portfolio by founding date and last funding event. Pull reports on any borrower whose business was incorporated before 2023 and whose industry is AI-adjacent (SaaS, data services, cloud consulting, automation). Contact those borrowers for an updated client list and revenue breakdown.

2. Restructure proactively. A stranded startup with positive unit economics but no VC path may still be a viable loan if terms are reset to match actual cash flow. Offer a payment deferral or interest-only period in exchange for a personal guarantee or additional collateral. The alternative is a full default that recovers pennies on the dollar.

3. Update your origination questionnaire. Add fields for “year business founded” and “date of most recent equity funding round.” If the answer indicates no funding since early 2023, require a written explanation of how the business will sustain itself without further venture capital.

4. Educate your risk team on the AI funding concentration phenomenon. The $250 billion that went to OpenAI and Anthropic did not appear from nowhere—it was redirected from the 200+ startups that would have raised Series A and B in 2023 and 2024. That money is gone. Lenders who treat AI as a monolithic growth story are missing the stranded cohort beneath the surface.

The Flip Side: A New Borrower Category

Not all AI-related small businesses are stranded. The concentration of funding has created a boom for companies that directly serve the large model builders and their hyperscale infrastructure needs. Data center electricians, cooling system contractors, fiber-optic installers, and GPU resellers are seeing explosive demand. Lenders should also target this cohort, but with care. Their revenue is tied to a small number of large customers, and any slowdown in model training spend would hit them hard.

For brokers, the opportunity lies in connecting stranded startups with turnaround financing—but only for those with genuine product-market fit that got caught in the narrative shift. A business with 80 percent gross margins and a sticky customer base that predates ChatGPT can be refinanced into a sustainable, slower-growth company. That is a viable loan if the lender understands the borrower’s new trajectory.

The Bottom Line

Pre-ChatGPT startups are not dead. But they are disrupted. For small business lenders, the distinction between a healthy borrower and a stranded one now hinges on a single date: November 30, 2022. That is the day the rules changed. Lenders who adjust their underwriting to account for the AI funding concentration will protect their portfolios. Those who ignore it will inherit a credit event that unfolds in slow motion, one missed payment at a time.

Operators, brokers, and ISOs who want to stay ahead should start the audit today. The signal is clear. The only question is whether you act on it before the defaults roll in.