Anthropic has filed its S-1 to go public on the Nasdaq. Trading could start as soon as next month, at a valuation somewhere between $1.5 trillion and $2 trillion. The Verge AI reports that the filing runs to hundreds of pages, and it’s the first real look inside the finances of a frontier AI lab. On Decoder, Nilay Patel walked through it with Joey Brookhart, an analyst at SemiAnalysis who co-leads the firm’s Tokenomics model.
The price is a big deal on its own terms. What matters more is the bet behind it. Wall Street is being asked to pay for AI that has barely spread past one use case, on the assumption that it’ll soon take over a large share of white collar work.
The Numbers Look Clean
Brookhart came away impressed. He called it “an awesome S-1” and said he didn’t see “much to nitpick in this filing.” What stood out to him:
- $70 billion+ in annualized recurring revenue (ARR), built in a very short time
- Gross margins, compute margins and free cash flow that all look like “what you’d want to see”
- A market rally in semiconductor, neocloud and compute stocks in the two days after the filing
That’s a sharp change in mood. Brookhart said the three weeks since Anthropic’s investor day had been “rough,” because some investors were let down by its September ARR numbers. The filing seems to have settled those nerves.
SemiAnalysis is worth taking seriously here. The firm built an outside-in model of Anthropic’s finances back in July. So this wasn’t a first look for them. It was a check on their own forecast.
The Real Question: Where Does $1 Trillion Come From?
Dario Amodei has said he can see a path to $1 trillion in revenue. Patel pushed on the obvious problem: if AI eats other businesses, who’s left to pay the bill?
Brookhart was frank about how narrow the market still is. The two leading labs together bring in about $120 billion in ARR, but the spending is highly concentrated:
- The top 1 percent of customers spend $50,000 to $100,000 per employee
- The broader Fortune 500 and Global 2000 “isn’t spending much”
- Anthropic is “almost a single vertical company,” and that vertical is coding
To grow from here, Brookhart pointed to cybersecurity, where eight-figure contracts can scale fast as incidents pile up, and to biotech. But he said the $1 trillion figure depends on moving “from AI augmentation to labor saving or replacement” in white collar work. He cited Microsoft’s employee buyout offers as an early example of that shift.
The Competition Problem
Patel pressed on what I see as the weakest part of the story. Anthropic doesn’t have a two-player market. There are other frontier labs and open-weight models. Plenty of CEOs don’t want to hand their data to OpenAI or Anthropic, and some law firms now run AI on their own servers.
Brookhart agreed that switching costs are low in parts of coding. He noted that OpenAI caught up in Q3. That’s a real risk for a company whose revenue sits so heavily in one category.
What Happens Next: The 1-3 Year View
The next big test is close. Brookhart pointed out that neither leading lab has been through a meaningful enterprise Q4 budgeting cycle, because their ARR only took off at the start of this year. This quarter’s budget decisions will show whether AI spending is spreading or staying in the coding niche.
Here’s what I expect over the next few years:
- 2026-2027: Enterprise budgets either broaden past coding or they don’t. Expect cybersecurity to be the second big category.
- 2027-2028: The jump from augmentation to replacement gets tested in finance, legal and operations. More announcements like Microsoft’s buyouts would signal that the thesis is working.
- Ongoing: Pricing pressure from open-weight and on-prem models caps margins on commodity work.
What Practitioners Should Do
- Don’t lock into one vendor. Low switching costs work for you as a buyer, so keep your coding and agent workflows portable.
- Benchmark against open-weight models for sensitive or high-volume work before you sign a big contract.
- Treat the Q4 budget season as your leverage. The labs need enterprise wins to back up these valuations, so negotiate hard.
The S-1 shows a strong business. Whether it’s worth $2 trillion depends on a future that hasn’t arrived yet. The full Decoder conversation is available at The Verge AI.