It’s not the model. It’s not the team’s fault. It is unit economics.
Venture capital invested a record $202 billion in AI start-ups in 2025, accounting for half of all global funding. But the math remains brutal: 90% of AI companies will fail, a rate significantly higher than the 70% seen among traditional technology startups. According to Alexander Kopylkov, a venture capital investor who focuses on long-term business fundamentals, this failure rate is not due to a lack of innovation, but rather to broken unit economics. “Anyone can create a demo,” he notes. “The survivors are the ones who can build a business.”
The combustion problem
Many Series A AI startups burn $2 to $5 for every new $1 in revenue. This burn multiple, a metric popularized by investor David Sacks, has become the crucial number that VCs are scrutinizing in 2026.
For comparison, the best-performing SaaS companies operate with burn multiples below 1.5x. The gold standard is 1x or less: Spend a dollar, earn a dollar.
Kopylkov breaks it down into first principles: AI startups face a structural cost problem that traditional software companies don’t have. “While a SaaS company spends 15-20% of revenue on infrastructure, AI companies often start at 40-50%,” he explains. “This gap must be closed, otherwise the company will go under.”
Infrastructure strain is not the only culprit. AI startups are also facing rising talent costs as machine learning engineers command salaries that dwarf traditional software roles. Add to this the constant need to retrain models, maintain data pipelines, and keep pace with rapidly evolving base models, and the cost structure becomes a burden.
What the survivors look like
Citing data from multiple VC surveys, Kopylkov finds that companies that achieve burn multiples below 1.5x share three characteristics: disciplined mindset, strict focus on product-market fit before scaling, and AI-powered operational efficiency.
The survivors have something else in common: corporate customers. Anthropic, one of the few AI companies with sustainable economics, generates 70-80% of its revenue from enterprise customers. The annual revenue rate rose from $87 million in early 2024 to $7 billion to $9 billion by the end of 2025, not through hype, but by solving compliance and security issues that large institutions have to pay for.
Kopylkov emphasizes that the company’s focus is not just about larger contracts. “Enterprise customers have longer sales cycles, but also lower churn, higher lifetime value and more predictable revenue,” he says. “This predictability allows you to plan, hire and scale without jeopardizing your runway.”
For founders, Kopylkov recommends a simple framework: Before you launch your next round, answer three questions. Is your combustion factor less than 2x? Do you have more than 18 months of catwalk experience? Are your gross margins above 50% or are they trending there quickly?
If the answer to any of these questions is “no,” investors will take notice in 2026. Due diligence has become more stringent and patience for ambitious forecasts is exhausted.
Consolidation is coming
The era of experimentation is coming to an end. According to a TechCrunch survey of 24 enterprise-focused VCs, 2026 is the year companies will start consolidating AI investments and picking winners.
Andrew Ferguson of Databricks Ventures put it best: “Today, companies are testing multiple tools for a single use case. As companies see real evidence of AI, they will cut some experimentation budgets, streamline overlapping tools, and roll those savings into the AI technologies they’ve delivered.”
In his view, this consolidation will accelerate through 2026 and 2027. The startups that survive won’t be the ones with the best pitch decks. They will be the ones with the clearest ROI.
For Kopylkov, this winning is inevitable. “If every startup claims to be AI-powered, the label becomes meaningless,” he says. “Buyers are getting smarter. They’re asking more sophisticated questions about what’s actually under the hood and whether the product delivers measurable value. The companies that can’t convincingly answer these questions won’t make it to 2027.”
The opportunity in the wreck
Despite the grim statistics, Kopylkov sees an opportunity. The 90% failure rate is not a reason to avoid AI, but a filter.
“The companies that come through are battle-tested,” he says. “They’ve proven they can efficiently acquire customers, retain them, and improve their margins over time. That’s exactly what you want to invest in.”
Kopylkov compares this shift to the dot-com era: lots of destruction, but the survivors like Amazon, Google and eBay defined the next two decades of technology. The pattern is familiar: irrational exuberance, painful correction, and then sustained growth based on real fundamentals.
The difference in 2026 is that investors are not waiting for the crash to demand fundamentals. They are demanding it now. The funding environment has shifted from “act fast and find out” to “show me the numbers.”
This is healthy for Kopylkov. “Capital discipline forces founders to think like entrepreneurs, not just visionaries,” he says. “The best companies that emerge from this period will have both.”
“2026 is a year where fundamentals come first, where capital rewards revenue growth, efficiency and real AI benefits – and punishes anything that represents an AI facade on old ideas.”
— Anders Ranum, Partner, Sapphire Ventures
For founders building AI companies today, the message is clear: the hype has opened the door for you. The economics of the unit will determine whether you stay.
