Financial Disclosures Ahead of IPO

As Anthropic approaches its initial public offering, financial disclosures are providing new insight into the economics of operating a major artificial intelligence lab, particularly regarding the proportion of revenue consumed by compute costs.

An updated S-1 filing released on Monday extended financial results through September 30, showing quarterly revenue reaching $17 billion, up from $11.5 billion in the previous quarter. However, operating losses also grew during this period due to rising infrastructure expenses.

Compute expenses accounted for 58% of company revenue in the third quarter, compared to 48% in the preceding period, indicating worsening margins at a time when leadership might seek to demonstrate improving efficiency to prospective investors.

Revenue Efficiency and Short-Term Commitments

A chart in the regulatory filing tracks revenue generated per dollar spent on compute, showing an increase starting in early 2025 that peaked at slightly over 2x before declining to 1.72x in the most recent quarter.

Image Credits:Anthropic S-1 filing

The margin shift stems largely from rapid demand growth that forced Anthropic to secure short-term compute commitments. The filing notes that $5.0 billion of compute was acquired through short-term arrangements at significantly higher rates than multi-year contracts, pushing the quarterly compute bill to $10 billion.

Reliance on Alternative Cloud Providers

Publicly reported agreements include substantial monthly payments to SpaceX alongside multi-billion-dollar contracts with neocloud providers such as Nscale, Akamai, and Volta, which offer flexible capacity at premium pricing compared to traditional hyperscalers.

The company expects infrastructure expenditures to remain its primary operating expense as multi-year cloud service provider arrangements gradually become available and additional investments are made.

While internal initiatives such as the "Anthropic Direct" project aim to expand proprietary compute capacity, rapid demand growth ensures continued reliance on alternative cloud providers.

Broader Infrastructure Vulnerabilities

With quarterly revenue growing rapidly, high compute expenses may remain secondary to investors, yet the situation underscores the heavy reliance of artificial intelligence laboratories on the broader data center infrastructure buildout.

Consequently, the financial outlook for Anthropic and similar laboratories remains closely tied to fluctuations in compute pricing and infrastructure availability.