Anthropic's $67B annualized run rate disappoints inflated market expectations, sending AI stocks lower

Anthropic's $67B Run Rate Falls Short of Market Hype

Anthropic announced a $67.2 billion annualized run rate at its investor day, representing a tenfold increase from a year prior. The figure disappointed investors who had bet on a larger surprise — expectations had shifted toward $90 billion or higher, fueled by speculation that Anthropic was sitting on undisclosed momentum.

The company has been fighting multiple revenue headwinds that likely constrained growth. Meta, a major customer, has begun using its own models internally and appears to be reducing reliance on Anthropic's API. An executive at Meta reportedly signaled the company could weaponize contract cancellation as leverage during Anthropic's IPO process, a veiled threat aimed at preventing a higher valuation that would exceed Meta's own.

The distillation tax

Model distillation — the practice of extracting and replicating proprietary AI outputs to train cheaper competitors — represents between 5 and 7 percent of Anthropic's revenue, translating to roughly $3.4 billion to $4.7 billion in annualized fraud. Anthropic deployed frontier technology to combat the practice and claims the company has successfully throttled distillation attacks. The figure is lower than some had anticipated, but remains economically significant.

The architecture of distillation at scale has grown sophisticated. Distillers operate through networks of third-party wrapper companies — SaaS products like AI-native document editors that position themselves as neutral infrastructure. A customer pays the wrapper a fee, which forwards queries to Anthropic's API and returns results. Distillers spin up thousands of accounts across these wrappers, making it difficult for either the wrapper operator or Anthropic to distinguish between organic usage and coordinated extraction. The wrapper company often remains unaware it is being used as a distillery, believing it has found an unusually engaged customer base, when in fact it has become a turnkey vehicle for model theft.

Anthropic chose not to pursue aggressive countermeasures like banning open-source model development or restricting API access. Instead, the company treated distillation as a technical problem to be solved through its own frontier capabilities. The choice reflects a deliberate trade-off: accepting near-term revenue loss to preserve long-term competitive positioning and avoid regulatory or reputational backlash.