Anthropic’s Best Financial Metric Is Also Its Biggest Risk

Dow Jones16:50

Along with OpenAI, soon-to-be-public Anthropic is one of two companies making the most advanced artificial-intelligence models. Its Claude models have become the first choice for enterprise AI applications like writing code, just as adoption is taking off. That is proving to be a significant advantage.

In September, I attended the Salesforce Dreamforce conference in San Francisco. I asked attendees about how they are implementing AI. Answers were all over the map, but two things were common to almost every discussion. First, this was the year everyone began to implement AI seriously. Second, even though Anthropic’s models are the most expensive, enterprises see them as the best and are willing to pay up.

Claude has quickly become the “Nobody ever got fired for buying IBM” of AI models—the default choice for risk-averse managers. Almost all the top firms in tech, banking, pharmaceuticals, and healthcare use Claude, according to the company’s initial public offering presentation.

These two areas of consensus are evident in Anthropic’s eye-opening IPO prospectus.

At the end of 2025, Claude began to improve rapidly, and Anthropic started 2026 with a ton of momentum. Fourth-quarter sales in 2025 were $2 billion, and by the third quarter of this year they had soared to $17 billion. Venture-capital firm Menlo Ventures estimated that enterprises spent $177 billion on AI services in the first nine months of 2026, up from $33 billion in all of last year. Anthropic sales grew even faster than that.

What’s more, in the last two quarters, the company had an adjusted operating profit, and for the first nine months of the year, it collected $5.6 billion in free cash flow. This stands in stark contrast to The Wall Street Journal’s reporting on OpenAI’s financials. According to Journal sources, in the third quarter OpenAI garnered about $10 billion in sales, but also recorded an $11 billion adjusted operating loss. The -109% operating margin was actually an improvement over -150% in the previous quarter.

By comparison, Anthropic looks like a lean, mean AI machine. It may be a reflection of the divergent go-to-market strategies the start-ups pursued beginning in 2023, when Anthropic had only $41 million in sales. OpenAI dominated on the consumer side with its ChatGPT app, monetized with premium subscriptions and advertising. Anthropic targeted enterprise applications—paid by usage—and Claude developed a reputation as the model built for business. That now looks to have been the better choice.

A trip down Anthropic’s income statement reveals the two big moving parts to the neat trick of hypergrowth coupled with profitability.

Anthropic’s usage-based pricing comes with a massive markup. In the first nine months of 2026, only 14% of sales went to cost of revenue, a nearly 600% markup. That Anthropic has this kind of pricing power is a reflection of its pole position in model capabilities.

But the cost of maintaining that status is further down the income statement in research and development—58% of revenue so far this year—which is the cost of training and testing new models.

The second of the 88 risk factors listed in Anthropic’s prospectus reads, “Our success depends in large part on our ability to continue to develop the most advanced and capable AI models.” High R&D spending has to remain a fixture.

Anthropic’s brand is “We’re the best, and you’re gonna have to pay up for it.” So the action turns to keeping Anthropic’s markup high to offset the cost of remaining the number one choice and retaining pricing power. Anthropic’s financial strength—an absurdly high gross margin—is also its greatest weakness because competition at the frontier is increasing. Anthropic used to only compete at the frontier with OpenAI. Now Meta Platforms and Alphabet have become more competitive too.

But the biggest threat to Anthropic’s premium pricing may be free “open-weight” models which enterprises can run themselves on their own cloud computing at a large savings, and which are rapidly closing the performance gap with Claude.

Anthropic has already lowered prices this year as a response to the competition. To keep its hypergrowth going, it may eventually be forced to trade it for a lower gross margin. To remain profitable, it will also have to try to get operating leverage on other expenses: sales, marketing, general and administrative, which together absorbed 31% of revenue in the first nine months of this year.

The other coming challenge is the results the 2026 AI adopters get. In 2027, leaders are going to take a hard look at their AI spending. They may decide to pull back if there aren’t tangible results. Historically, new general-purpose technologies like AI initially reduce productivity until enterprises can retrain workers, build new applications on top of them, and rebuild business processes around them. The Apple II was released in 1977, but the productivity gains from PCs didn’t appear in the macroeconomic data until the 1990s.

Anthropic is holding itself out as an AI-adoption model for others to copy. Revenue per employee was $6.6 million in the last 12 months, up from $1.4 million the year before. That is the kind of productivity gain Anthropic needs its customers to also see.

Both Anthropic’s big challenges—pricing power and its customers seeing return-on-investment—come due next year.

At the request of the copyright holder, you need to log in to view this content

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment