Anthropic’s $16.8 billion quarter faces a tougher IPO test

Inference is producing profit

Anthropic’s adjusted operating margin for the year through the third quarter was 11%, according to The Information. The measure excludes major costs such as stock-based compensation, but still points toward profitability.

The Financial Times reported that Anthropic has now had two consecutive quarters of adjusted profit. It also generated positive cash flow: in the third quarter, every dollar spent on computing brought in $1.70 of revenue.

That is meaningful evidence that model companies can make money, but the caveats matter. Training and staffing remain expensive, while open-weight models could put pressure on closed products such as Claude.

Blocking distillation comes at a cost

Anthropic estimates that distillation accounts for 5% to 7% of its revenue. Other companies use Anthropic’s AI to extract information about how it works and train their own models. Anthropic has repeatedly accused Chinese AI labs of copying its models this way, and recently introduced effective safeguards.

The company has told investors that distillation revenue is now likely to decline, affecting results in the current quarter. In the longer term, blocking the practice should make it harder for competitors to reach Claude’s quality.

I think this is a revealing trade-off: a revenue source the company says it wants to shut down has been material enough to affect near-term results. The announcement does not say how much of that 5% to 7% came from distillation, or how quickly the safeguards will change the economics.

The IPO case is missing a growth number

Anthropic plans to go public in November, before Thanksgiving, according to The Information. The offering will test investor appetite for the AI boom. In May, the company’s private valuation reached $965 billion, up from less than $20 billion two years earlier. Some bankers have floated a possible IPO valuation of $2 trillion.

But reports say the revenue figures disappointed some potential investors. Some now expect Anthropic’s advising banks to lower the valuation, possibly below $1.5 trillion. The company did not disclose its September revenue run rate or a full-year forecast; that run rate was reportedly $65 billion in July.

The wider setting is difficult: inflation remains high, borrowing costs are rising, public sentiment toward AI is more negative than positive, and unresolved cybersecurity risks pose an operational threat. OpenAI, Anthropic, Google and others have also been locked in a price war for months, while the quality gap between closed and open models is narrowing.

My guess is that the IPO debate will turn less on whether Anthropic can show profit than on whether investors can see growth continuing at a pace that justifies the valuation. People are likely to use AI more as better models spread into everyday office work, but that does not settle how much of the value Anthropic can keep.