Anthropic Cuts AI Model Prices, Faces Heavier Cost Burden and Swings to Q3 Loss
Summary
- Anthropic said about $600 million in revenue growth was offset by price cuts for AI models and contract discounts.
- It also said computing and infrastructure costs surged about 77% to $9.86 billion from $5.56 billion over the same period, resulting in a third-quarter operating loss of $567 million.
- In an amended filing ahead of its IPO, the company said total lease and purchase commitments reached about $895.4 billion, with future fundraising capacity emerging as a key challenge.
Forecast Trend Report by Period
Anthropic has lowered prices for its Claude models to attract customers, but a surge in computing costs is increasing pressure on profitability.
Decrypt reported on November 3 that Anthropic said in an amended filing with the US Securities and Exchange Commission that price cuts for its AI models and contract discounts offset about $600 million in revenue growth. Lower pricing weighed on results even as usage expanded.
Computing and infrastructure costs jumped about 77% over the same period, rising to $9.86 billion from $5.56 billion. Computing and infrastructure spending per dollar of revenue increased to 58 cents from about 48 cents in the previous quarter. Those costs include not only customer service but also computing resources used for AI model research, training and internal operations.
Anthropic recently lowered token pricing for Claude Opus 5.5 versus its previous model, and in October introduced Haiku 5.5, a lower-cost model designed for high-volume workloads. The company said the cost per task for Opus 5.5 fell about 40% from the earlier model. Its strategy is to expand usage through lower prices and better performance.
But profitability deteriorated as infrastructure procurement costs climbed rapidly. Anthropic typically secures computing resources 12 to 18 months before they are used, leaving a significant share of those costs difficult to adjust in the short term. In the third quarter, it spent about $5 billion on short-term computing contracts, which are more expensive than long-term agreements.
As a result, Anthropic swung to a third-quarter operating loss of $567 million after reporting $974 million in operating profit in the second quarter. Demand for AI services is rising quickly, but the infrastructure investment needed to support that growth is limiting profitability gains.
The companys longer-term financial burden is also coming into focus. In the amended filing released ahead of its initial public offering, Anthropic said total contractual commitments related to leases and purchases amounted to about $895.4 billion. Most of that is tied to computing and infrastructure costs, making its ability to finance future large-scale investment a key challenge.
Anthropic filed a registration statement last month for a Nasdaq listing under the ticker ANT. The market is watching whether the company can translate rising usage into improved profitability as competition in AI model pricing intensifies.
