The jump: Revenue went from $787M in Q2 2025 to $11.6B in Q2 2026. Claude Code was the main reason for the explosion in growth in early 2026.

The catch: AI coding is not a monopoly. Open weight models and other labs can do it too. But Anthropic is still leading in that category.

Revenue versus compute: Compute and infrastructure expense fell from 251% of revenue in Q1 2025 to 48% in Q2 2026.

First operating profit: Operating income was $974M in Q2. Adjusted operating income was $2.24B.

Margins: Operating margin is 8% on a GAAP basis. Adjusted operating margin is 19%.

The scary number: The GAAP net loss was $80.7B in the first half. This is a paper loss, not the cash in and out of the business.

Where it came from: $72.4B of convertible notes were revalued, plus $8.7B from other mark to market items. It's similar to how stock based compensation hits GAAP net income.

Adjusted net income: $2.98B in Q2 and $2.7B in the first half. That strips out stock pay and the revaluations.

Cash flow: Operating cash flow was $3.9B in the first half of 2026. The S-1 doesn't give a free cash flow number, but the math says $3.4B.

Capital light: Anthropic isn't building the data centers. It has contracts for future compute bills instead. Capex was only $482M in the first half.

The balance sheet: $131.7B of cash and short-term investments, after raising $141.1B to date. That's before whatever it raises in the IPO. A war chest like that means it can ride out the waves if raising capital ever gets harder.

Big spenders: 11,900+ customers spend $100K+ a year, up from 950+ the year before. 1,900+ spend $1M+, and about 250 spend $10M+ (trailing 12 months).

Net revenue retention: About 900%. Existing customers spend about 9x what they did a year ago, because the models got far more capable. We probably won't see 900% again next year, but I expect its top spenders to keep getting bigger.

Concentration: About 25% of revenue came from the top five customers in the first nine months of 2026.

Alphabet ($GOOGL): At least $114.5B, through July 2033.

Broadcom ($AVGO): About $161B of TPU leases (the Google chips), on 5 year leases.

SpaceX ($SPCX): Up to $120.5B if it runs through 2029.

Amazon ($AMZN): At least $110B, through April 2036, for AWS and its Trainium chips.

Microsoft ($MSFT): At least $44.9B, through May 2033.

Neoclouds: Nscale and Lambda have huge contracts running through 2033 and 2034. They aren't public companies.

If Anthropic does well: It means very good things for Nvidia ($NVDA), Broadcom, AMD ($AMD), Amazon, Microsoft, Alphabet, CoreWeave ($CRWV) and SpaceX.

The S&P 500: Some of its largest holdings are Nvidia, Microsoft, Amazon, Alphabet, Broadcom, Meta ($META) and Micron ($MU). The link isn't in revenue and fundamentals today. It's that a lot of their future growth is planned to come from AI labs like Anthropic.

The IPO: Am I buying it? Definitely not. Historically, IPOs have been the worst time to buy a stock. Warren Buffett has talked about why: with all the coverage, incentives and fundraising happening, the odds an IPO is the best thing to buy are practically zero.

Long term: I'm very open to adding Anthropic to my portfolio. I think there will be plenty of opportunity over the next couple of years, and you do not have to rush.

Volatility: The AI trade is the most volatile thing in the entire stock market. Some people think it's a bubble. In the macro view, I think they're wrong. AI as a technology is very real, and it will be transformational over the coming decades.

What moves it: The government debacle earlier this year. The constant threat of open weights. Talk that growth will slow. Any of these could send the stock down a lot in a day. It could also trade crazy at times, like the inversion of the software selloff earlier this year, when a lot of that money would probably flow into companies like Anthropic.

Earnings This Week

Dividend News This Week

Dividend Dates This Week