Venture capital — the business of buying equity stakes in young, private companies — has always rewarded patience. The AI cycle is testing that. Growth that once took a decade now takes quarters, and the investors writing the checks say the hard part is no longer spotting opportunity. It is staying disciplined when everything is moving too fast.
The scale of the moment
The numbers are hard to overstate. Global venture firms invested roughly $300 billion across about 6,000 startups in the first quarter of 2026, an all-time high, with AI absorbing close to 80% of that capital. Median pre-money valuations — the value placed on a company before new money goes in — nearly doubled in a single quarter, jumping from about $30 million to roughly $70 million. And the money is highly concentrated: just three companies — OpenAI, Anthropic and xAI — drew about two-thirds of all AI funding in the quarter.
"Cocktail napkin math"
In a TechCrunch discussion, M13 co-founder Carter Reum and Basis Set Ventures partner Chang Xu described how they try to think through that noise. Reum argued that AI growth rates are genuinely without precedent — pointing to products reaching billions in revenue in months — but that not every deal justifies an astronomical price. The defense, in his telling, is unglamorous: "cocktail napkin math," a quick sanity check of whether a company's assumptions hold up before the fear of missing out, or FOMO, takes over.
That fear is the defining emotion of the cycle. When a competitor is closing a round in days, due diligence — the investigation an investor does into a company's finances, technology and team before committing — can feel like a luxury. AI tools are now used to compress that work, cutting diligence timelines from months to days. The promise is speed without losing depth; the risk is that fast diligence becomes shallow diligence, and capital flows toward momentum rather than substance.
Where the moats are
The investors' second concern is defensibility — what protects a startup once it succeeds. In past cycles, founders mostly competed against other founders. Now they also face the largest companies in the world, which have more data, distribution and computing power. That changes where a "moat," or durable competitive advantage, can be found. Reum and Xu point toward regulated or hard-to-enter markets such as healthcare and emergency services, where "friction as a moat" keeps the giants at bay. Industry analysts make a similar case: the real protection for vertical AI companies is workflow depth, data rights and compliance scaffolding, not model sophistication alone.
They also distinguish between two kinds of markets. In "velocity" markets, the winner is whoever ships and iterates fastest. In "depth" markets — biotech, for instance — hard technical problems take time regardless of AI hype, and speed cannot substitute for science. The implication is that no single playbook fits every bet.
The case for caution
That concentration of capital has drawn comparisons to past manias, and some prominent voices have warned of correction risk tied to AI overvaluation. None of this is investment advice, and the investors quoted are careful not to predict an outcome. The more useful takeaway is a way of thinking. In a market this fast, the edge may belong less to whoever moves first than to whoever can still do the arithmetic before signing.
