Accel raises $5 billion for AI startups, intensifying the race for late-stage rounds

Accel has raised almost $5 billion for a new cycle of big bets on artificial intelligence. Venture capital increasingly looks less like classic startup financing and more like a race of infrastructure budgets, in which a single fund is ready to write $200 million checks and compete not with ordinary VCs but with mega-funds, corporations and sovereign investors.

The new capital comes in two parts: $4 billion goes to Leaders Fund V, and another $650 million to a sidecar fund that lets the firm increase its stake in the strongest deals, primarily within its existing portfolio. According to Accel's partners, the main fund is built for roughly 20-25 investments in late-stage companies worldwide. The average check is around $200 million. This format is not meant for pre-seed experiments, but for companies that have already found product-market fit, are growing revenue quickly and need capital on the scale of infrastructure projects.

The fund's logic is clear from its latest bets. Accel invested in Anthropic last year at a valuation of about $183 billion. The company then closed a new round at roughly $380 billion, and offers implying a valuation of about $800 billion have appeared on the market.

The Cursor story is similar: in June 2025 the firm invested in the AI code editor at a $9.9 billion valuation, in November the company raised a new round at $29.3 billion, and in March 2026 it was discussing raising capital at roughly $50 billion.

For Accel these are not just successful deals but a demonstration of why a large fund is needed: if you back one of the cycle's leaders, you need to be able to keep betting rather than watch from the sidelines as others buy up the round.

It is not only about these two big names. Accel's AI portfolio also includes Vercel, n8n, Recraft and Code Metal. In March the firm, together with the Google AI Futures Fund, launched the Atoms AI program for early-stage teams looking for unoccupied niches in enterprise AI. The strategy is therefore twofold: catch future winners as early as possible, while keeping the ability to pour hundreds of millions into them once the market has proven demand and growth speed. In practice this brings venture capital closer to private equity and infrastructure financing.

This deal shows clearly how the whole market has changed. In the first quarter of 2026, startups worldwide raised a record $297 billion, 2.5 times more than the quarter before.

Andreessen Horowitz raised $15 billion, Thrive Capital closed more than $10 billion, and Founders Fund is finalizing a $6 billion fund. Against this backdrop Accel no longer looks like an exception, although just a few years ago raising $5 billion would have been an event on the scale of a single tech giant. The middle of the venture market is shrinking: large players are moving down the stages, while strong AI companies sometimes skip the usual funding steps and grow from seed to multibillion-dollar valuations in a year and a half.

Accel draws on historical experience. The firm was founded in 1983 and in 2005 invested $12.7 million for 10% of Facebook, a stake worth $6.6 billion by the IPO. But the new AI cycle is different in that the size of the capital has become part of the product: to build models, data centers, hardware, robotics or defense tech, it is no longer enough to be simply a good venture investor. You need to make decisions fast and just as fast find hundreds of millions for the next rounds. That is why Accel separately talks about future bets at the intersection of software and hardware, from robotics to infrastructure for AI data centers.

The main takeaway is simple: the AI market has entered a phase in which the winners are not only those with the best products but those with the best balance sheets. For founders, this means access to nearly industrial-scale money if the company shows a rare growth rate. For investors, it means a higher profit ceiling but also a more painful risk of overpaying at the peak of the cycle. And for the market as a whole, it is further confirmation that venture capital in AI has finally turned from a hunt for startups into a fight for infrastructure champions.

*Meta is recognized as an extremist organization and banned in Russia.