Record capital is flowing into AI, but fewer startups are getting any of it
If you’ve been following the news, you might think AI is either taking over the world or about to collapse. The truth is simpler, and more useful to understand.
AI attracted $258.7 billion in global venture capital in 2025, according to the OECD. That’s 61% of all the money invested in startups worldwide. The AI funding boom is real.
But there’s a catch most headlines miss: while the dollars went up, the number of deals went down. In Q4 2025, global VC recorded only 9,844 deals — the lowest quarterly count since early 2020. More money. Fewer winners.
The market isn’t collapsing. It’s concentrating.
The Numbers Behind the Shift
Think of it like a funnel that got narrower at the bottom.
In the first half of 2025, just 11 companies captured more than a third of all global venture capital — roughly $70 billion out of $205 billion total. OpenAI and Anthropic alone accounted for 14% of all global startup investment.
Meanwhile, the number of active VC funds globally fell from 4,430 in 2022 to just 823 in 2025. That’s an 81% drop. Fewer investors writing checks — and the ones who are writing them are writing much bigger ones, to fewer companies.
According to Kopylkov, this is a market correction, not a crash. “When capital concentrates, it usually means the market has gotten smarter — not scared. Investors are no longer paying for the idea of AI. They’re paying for proof.”
Two Companies Tell the Whole Story
ElevenLabs, a voice AI company, ended 2025 with $330 million in annual revenue — up 175% from the year before. It generates roughly $825,000 in revenue per employee. In early 2026, it raised $500 million at an $11 billion valuation. Its enterprise clients include Deutsche Telekom, Revolut, and Cisco. Customers stay, pay more over time, and refer others.
Builder.ai raised $445 million from investors including Microsoft. It was valued at more than $1 billion. In May 2025, it filed for bankruptcy. Investigators found that the company’s claimed $220 million in revenue was actually $55 million. Work marketed as AI-generated was being done manually by hundreds of engineers. No real product moat. No honest metrics.
Same market. Opposite outcomes.
Kopylkov breaks it down simply: “ElevenLabs built something people keep paying for. Builder.ai built a story. In 2021, both could get funded. In 2025, only one of them survives.”
What Investors Are Actually Asking Now
The checklist has changed. In 2021, a great pitch deck and fast growth were often enough. Today, investors want to see the numbers behind the growth.
The most important ones — explained plainly:
- Burn Multiple: How much money are you spending to earn each dollar of new revenue? Below 1x is excellent. Above 3x, most investors won’t take the meeting.
- CAC Payback: How long does it take to recover the cost of acquiring one customer? Under 12 months is healthy. Under 6 months is outstanding.
- Net Dollar Retention (NDR): Are your existing customers spending more with you over time? 120% or above means your revenue grows even without new customers.
Citing benchmarks from Bessemer Venture Partners, Kopylkov notes that top AI companies are now generating over $1 million in revenue per employee — roughly four to five times the traditional software industry average.
For founders, Kopylkov recommends starting with one simple question before raising capital: “If we stopped acquiring new customers today, would our revenue grow, shrink, or stay flat?” If the answer isn’t “grow,” the fundamentals need work first.
The Pattern Shows Up Everywhere
This isn’t a uniquely American story. The same dynamic is playing out globally. In Latin America, venture investment recovered to $4.1 billion in 2025 — a 13.8% increase. But in Brazil, the region’s largest tech market, the single biggest deal of Q3 2025 represented nearly a quarter of the country’s entire quarterly total. More recovery at the top. More pressure at the bottom.
In his view, the market will continue thinning out through 2026 and beyond — leaving behind companies with genuine unit economics, real customer retention, and products that solve actual problems rather than chasing trends.
The Takeaway for Founders
The AI funding wave hasn’t ended. It has changed shape. The companies getting funded in 2026 aren’t necessarily bigger or more ambitious than those getting rejected — they’re more honest about their numbers, more disciplined about spending, and more focused on keeping the customers they already have.
That’s always been what good business looks like. AI just made it easier to see who had it and who didn’t.
