Anthropic Hits $17B Q3 Revenue But Sinks Back to Loss on Massive Compute Costs
Anthropic just revealed a $17 billion quarter ahead of its IPO, but a $567 million operating loss shows the brutal cost of AI compute.
Business · Source: Reuters
What happened
Anthropic amended its IPO filing to include third-quarter numbers. Revenue surged to $16.96 billion, up from $11.56 billion in the second quarter. But the AI giant slipped back into the red. It posted a $567 million operating loss after showing a $974 million operating profit just three months prior.
The culprit is compute. Anthropic spent $9.86 billion on compute and infrastructure in the third quarter. That eats up 58 percent of its revenue. The company had to buy $5 billion in short-term compute capacity at a premium because customer demand outpaced its long-term contracts.
The filing also exposes the staggering scale of Anthropic's future. The company has locked in $895.4 billion in purchase and lease commitments. These deals tie Anthropic to Amazon, Google, Microsoft, and Broadcom for years to come. About $69.8 billion of that bill comes due in the next twelve months alone.
Key facts
- $16.96 billion — Anthropic's third-quarter revenue, up from $1.38 billion a year earlier
- $567 million — Operating loss in the third quarter, driven by expensive short-term compute
- $895.4 billion — Total future purchase and lease commitments, mostly for compute and infrastructure
- $21.9 billion — Non-cash expense from marking convertible notes to a higher company valuation
- 11,900 — Number of customers generating at least $100,000 in trailing 12-month revenue
Why it matters
The cost of building frontier AI models is accelerating faster than revenue. Anthropic grew its revenue by more than 1,100 percent year-over-year, yet its operating margins still compressed. For anyone building AI products, this proves that compute constraints are real and expensive. When demand spikes, even the biggest players have to pay a massive premium for short-term server access.
This filing sets a new baseline for the AI infrastructure boom. Anthropic is shifting toward Anthropic Direct, a model where it leases chips and hosts them in dedicated facilities instead of relying purely on cloud providers. This signals a broader industry move where top AI labs bring infrastructure in-house to control costs. It also tests public market appetite for businesses that require nearly a trillion dollars in future capital commitments just to stay competitive.
For builders
Prepare for compute price volatility
Anthropic took a massive margin hit buying short-term compute. Startups relying on spot instances or short-term cloud contracts will face similar price gouging. Lock in your compute needs early or build graceful degradation into your products.
Enterprise AI adoption is accelerating
Anthropic now has over 11,900 customers spending at least $100,000 annually. This is up from just 1,500 at the end of 2025. B2B founders have a massive opportunity to build tooling and integration layers for these high-paying enterprise buyers.
Cloud marketplaces drive major distribution
Sales through cloud provider marketplaces accounted for 44 percent of Anthropic's revenue over the last nine months. If you build B2B software, listing on AWS, Google Cloud, or Azure marketplaces is no longer optional. It is where the budgets are.
My take
You cannot build a frontier AI company without a blank check. Anthropic is printing money with $17 billion in a single quarter, but they are burning it on compute just as fast. The $895 billion in future commitments shows this is a game only tech royalty can play, and the upcoming IPO will prove if Wall Street actually has the stomach for it.
Original reporting: Reuters. This is my rewrite and opinion.