Can AI labs ever turn a profit? Joachim Klement
QQQ•Anthropic’s revenue rose to $4.6 billion in 2025, but its operating loss widened to $8.06 billion as it spent $7.3 billion on compute and infrastructure. Falling token prices and other indicators are raising questions about whether AI labs can grow usage enough to become profitable.
1. Anthropic’s widening losses
Anthropic’s IPO prospectus showed revenue jumped nearly 12-fold to $4.6 billion in 2025, while spending on compute and infrastructure reached $7.3 billion. Its operating loss widened from $2.98 billion in 2024 to $8.06 billion in 2025.
2. Token prices and demand
The average price of large language model tokens, as estimated by the LLM Token Expenditure Index, has fallen more than 40% since June 30. AI firms argue lower costs can spur enough additional use to raise overall consumption, but the article says there is little evidence so far that increased usage is offsetting price declines. The index is new and may miss parts of the market or differences between models.
3. Questions about valuations
OpenRouter’s share of requests sent to Anthropic models fell to 2.6% from 5.2% at the end of June and a May peak of 6.4%. The article also points to uncertainty around AI firms’ profitability and valuations, including SoftBank’s financing against its OpenAI stake and forecasts for US technology companies’ operating cash flow.
