There is a strange paradox emerging in venture capital. AI may become one of the greatest investment opportunities of our generation, but the more convinced investors become of that outcome, the more concentrated the venture market appears to become. According to PitchBook data cited by Axios, OpenAI and Anthropic alone reportedly attracted more than 60 percent of all venture dollars invested in U.S. startups during the first half of 2026. These are, of course, exceptional companies, but the number still illustrates how dramatically the market is changing. Venture capital has always been a power-law business in which a few companies produce most of the returns. What feels different today is that concentration is no longer happening only inside individual portfolios. Increasingly, the entire market wants exposure to the same small group of companies.

That creates an uncomfortable possibility: if the AI bet fails, venture capital may have accumulated enormous concentration risk. If it succeeds, the industry may become even more concentrated. Part of this development has little to do with AI itself and much more to do with the structure of venture capital. Funds have become much larger, and the bigger the fund, the bigger the outcome it needs. For a $300 million fund, a company selling for $2 billion can still produce a meaningful return. For a $10 or $15 billion fund, the same exit barely moves the needle. These funds increasingly need companies capable of becoming worth $100 billion, $300 billion or more, and AI happens to be one of the few markets where outcomes of that magnitude currently appear plausible.

From Picking to Access

This changes the investment decision itself. Traditionally, venture investing is supposed to be about picking: Which team is strongest? Which product is most differentiated? Which market will become large, and what price appropriately reflects the opportunity and the risk? In parts of today’s AI market, the decision increasingly appears to be about access instead. The relevant question becomes less “Is this the right price?” and more “Can we afford not to be in this company?” If OpenAI, Anthropic or another frontier AI company eventually becomes one of the most valuable companies in the world, then getting exposure at almost any earlier valuation may prove to have been rational.

The problem appears when many investors reach the same conclusion at the same time. If hundreds of funds independently decide that they must own the same handful of companies, individually rational behavior creates extreme concentration at the market level. The underlying assumption is not simply that AI will become a huge market. That seems increasingly difficult to dispute. The much more important assumption is that a very large share of the economic value created by AI will remain with today’s frontier model companies.

AI Will Be Huge. But Who Captures the Value?

This is where I think the investment thesis becomes more interesting. The biggest risk to the frontier labs is probably not that AI disappears, but that intelligence itself becomes cheaper. Open-weight models are improving quickly, prices are falling and the differences between leading models can narrow surprisingly fast. That does not mean open source will destroy companies like OpenAI or Anthropic, but it could materially change the economics of the market.

If models become increasingly interchangeable, the long-term moat may not primarily sit in the model itself. It may shift toward distribution, enterprise relationships, proprietary data, developer ecosystems, infrastructure or deep integration into customer workflows. For investors, that distinction matters enormously. A valuation of several hundred billion dollars requires more than extraordinary growth; it also assumes that the company will capture an extraordinary share of the future value pool. AI can therefore become one of the largest technology markets in history without every layer of the stack becoming equally valuable. The important question is not simply how big AI will become, but where the value will ultimately accrue.

The Concentration Flywheel

Now consider the optimistic scenario. The major AI companies eventually produce enormous IPOs, LPs finally receive the distributions they have been waiting for and venture capital’s liquidity problem begins to disappear. That sounds like great news, but there is a second-order effect. If the largest venture firms own disproportionate stakes in the biggest AI winners, they will also generate a disproportionate share of those distributions. And when LPs decide where to invest that money next, the obvious candidates are the managers who just demonstrated that they could access and profit from the biggest winners.

This creates a powerful concentration flywheel. Large funds can write larger checks, which gives them better access to capital-intensive and highly sought-after companies. Successful exits generate distributions, distributions attract more LP capital, and the next fund becomes even larger. AI could therefore solve venture capital’s liquidity problem while simultaneously concentrating more power in a relatively small number of investment platforms.

A Barbell Market for Startups and VCs

The same dynamic could emerge among startups. A $15 billion venture fund needs very different outcomes from a $300 million fund, which means more capital competes for a relatively small number of companies capable of producing enormous returns. The result could be a barbell market in which a small group of startups has almost unlimited access to capital, while many genuinely good companies struggle to raise because their potential outcome is simply not large enough to matter for a mega-fund.

Smaller VC firms could face the same pressure. If LP capital increasingly flows toward a few established brands, emerging managers may find it even harder to raise their first or second fund. In that sense, the winner-takes-most dynamic we are already seeing among AI companies could eventually appear among venture firms themselves.

What AI Is Really Changing

That is why I think the AI boom is interesting beyond AI itself. It is not only testing which companies can build the most valuable models or applications; it is also exposing how venture capital itself has changed. Funds are larger, liquidity has become more important, and the biggest platforms increasingly need extraordinary outcomes to justify their size. AI fits that model almost perfectly, which is precisely why so much capital is flowing into so few companies.

The question for the next few years is therefore not only whether the AI boom will produce extraordinary returns. It is whether those returns will make venture capital stronger as an asset class, or simply bigger, more concentrated and increasingly dependent on a smaller number of extraordinary winners.