By , a tech columnist at Intelligencer  Formerly, he was a reporter and critic at the New York Times and co-editor of The Awl.

Photo-Illustration: Intelligencer; Photo: Getty Images

You might say that the biggest AI start-ups are led by professionally unreliable narrators who theorize about the future of their companies — and their industry, and humanity in general — with a variety of clear but sometimes conflicting biases. This can make it somewhat hard, from the outside, to figure out which vision investors are banking on, beyond a general fear of missing out. Mass labor automation? AI-assisted research? Mainstream searchlike products that could unseat Google? Digital pals that occasionally sexualize young users or tell you how to kill yourself?

Lately, though, there have been a few shifts in the industry. One is vibey: A slightly underwhelming GPT-5 release, accompanied by a change in rhetoric from some AI leaders, has shifted the conversation away from acceleration and superintelligence and back toward nervous talk about bubbles. The other has shown up in deals: After Google beat out OpenAI to acquire (or acqui-hire) AI software development start-up Windsurf, both OpenAI and xAI are reportedly stalking AI coding-assistant start-up Cursor, which rebuffed OpenAI earlier this year but may now be exploring deals to license data; meanwhile, Nvidia, which has made enormous amounts of money selling hardware to AI firms, just acquired an AI coding start-up of its own. Both of these deals were signals that the industry is refocusing its attention on AI coding — a curious early use case for LLMs that has become one of the first clear and monetizable product categories for the technology. And some eyebrow-raising numbers from Anthropic this week prove that investors are onboard, too:

Anthropic has completed a Series F fundraising of $13 billion led by ICONIQ. This financing values Anthropic at $183 billion post-money… The investment reflects Anthropic’s continued momentum and reinforces our position as the leading intelligence platform for enterprises, developers, and power users.

One notable thing about this haul is its sheer size. It shows that, despite the wobbles and anxiety across the tech industry, companies with their own competitive models — and there are only a few — still spur massive interest from investors. Just as telling, though, is its relative size. OpenAI is reportedly negotiating share sales at a $500 billion valuation, up from $300 million just a few months ago, based on approximately $20 billion in annualized revenue and more than 800 million weekly users for ChatGPT. Anthropic, on the other hand, is reportedly running at about $5 billion in yearly revenue, up dramatically from January of this year, when it was just $1 billion. Anthropic doesn’t report user numbers, but they’re comparatively minuscule: According to Andreessen Horowitz and Similarweb, Claude, the company’s AI chatbot, is the sixth most-visited AI service on the web, with around a hundred million monthly visits — not unique visitors — and doesn’t crack the top 50 in the App Store. Semrush, which also tracks website visits, claims Claude gets about 3 percent of ChatGPT’s traffic.

Web visits are a flawed and incomplete metric for companies like this, which users interact with in a variety of ways: in apps, with APIs, and through other services that pay to use their models. But they help tell a pretty clear story about how investors are valuing AI companies these days. ChatGPT is already a household name, with most of its revenue coming from a few million paid subscriptions, while Google, Meta, and xAI are betting on mass adoption by bundling AI tools into already-popular products. Their valuations (and stock prices) reflect that investors believe these stories, and to some extent the belief that one of these firms might generally win, or at least win some major consumer market. Anthropic isn’t really in that race. But it is the clear favorite among customers who use AI to write software and has focused its efforts on cultivating those customers’ business. Considering that this is one of the few obviously lucrative case uses of AI, it’s a good place for Anthropic to be.

A lot of AI coding start-ups, some valued at billions of dollars on their own, are effectively wrappers around Anthropic’s models, paying hefty fees to access them. After a few years of vague, disorienting rhetoric around how AI companies might pay back the hundreds of billions of dollars they’ve raised, investors seem to be converging on a narrower thesis. They might not think they’re building God. But they might think they can automate part of the tech industry itself.

See All
How Investors Think AI Will Actually Make Money