How the escalating AI token costs force new Accel funding rules

This article summarizes an episode of Sourcery with Molly O’Shea’s video series featuring Arun Mathew, partner at Accel.

Arun Mathew, partner at Accel/ Photo credit: Arun Mathew

Arun Mathew, a partner at Accel, argues that late-stage AI deals now deliver the returns venture investors once expected from early bets.

This shift changes how investors price deals, how founders choose their backers, and which companies gain leverage as AI reshapes how software is distributed.

AI investing now needs range and discipline

The AI boom has made it easy to mistake available capital for good judgment. When every category looks promising, a loose investing plan becomes expensive quickly.

Instead of backing every AI trend, Accel filters the market by doing these steps:

• Study the whole AI market before committing, tracking everything from chips and AI cloud providers to labs, apps, and the companies that install these systems for customers.
• Use stage flexibly, writing early checks when a category is forming and growth checks when a company’s results are rising.
• Invest globally only where local relationships create access, such as in Silicon Valley, London, Bangalore, and Israel-linked security networks.
• Show conviction by investing early or writing a large check, rather than treating every popular AI category as a bet.
• Keep the portfolio narrow enough to understand how companies depend on each other, allowing the firm to act before prices rise.

As Accel’s partner, Matt Weigand explains, “Accel is pursuing the AI opportunity across every layer of the technology stack… We don’t believe in a blanket strategy and just spraying and praying.”

This connected approach raises the standard for investment committees. A firm might have the capital to write a billion-dollar check, but that power only matters if they walk away from similar companies that lack control over their market.

Founders are feeling this shift, too. Committed investors can support a company across funding rounds, geographies, and public or private markets. While this speeds up financing, it makes founders dependent on a smaller group of capital partners.

AI recommendations are becoming a sales path

For these capital partners, disciplined investing depends on finding where a company can shape customer demand. AI-driven product recommendations are becoming the battleground.

Instead of humans searching for software, AI models are guiding users toward tools, especially in developer workflows. Clear documentation, simple APIs, and easy setup now determine whether an AI agent recommends a product over a competitor.

Products must now be understood, used, and recommended by AI tools inside a task.

Mathew highlights this shift, “In the last era, it was search engine optimization. In this era, it’s AI optimization. These products are now the distribution mechanism for all the downstream products and sarervices that you can use, and they tend to prioritize the tools that have the best developer and user experience.”

Some tools now sit where agents make choices

Token use is outgrowing compute plans

Companies need fewer security tools they can trust