Greetings,

The race between Anthropic and OpenAI is often framed as a battle of intelligence, but right now, it's a battle of the balance sheet. While Anthropic is seeing a staggering revenue surge, new financial projections show just how expensive it is to keep that momentum going.

Sri Muppidi reports on Anthropic's revised 2025 outlook, where the company lowered its gross profit margin projection to 40%. While that is a massive jump from the negative 94% margin it saw in 2024, it's a notable 10 percentage point drop from earlier, more optimistic expectations. The culprit? The sheer cost of running these models for customers—known as inference—which came in 23% higher than anticipated.

Why it caught my eye:

• The Efficiency Gap: Anthropic's projected 40% gross margin suggests it may be operating less efficiently than OpenAI, which is eyeing a 46% margin for 2025 despite supporting a massive base of nonpaying users.
• Explosive Revenue vs. Intense Burn: Anthropic is on track for $4.5 billion in revenue for 2025—a nearly 12x increase from 2024—yet it still expected to lose about $5.2 billion last year.
• The Infrastructure Pivot: To stop the bleeding, Anthropic is pivoting toward hardware control, agreeing to buy $21 billion worth of Google's TPUs to rein in the computing costs that are currently eating their margins.

This is a fascinating look at the "hidden" costs of the AI boom and a reminder that even $4.5 billion in revenue doesn't guarantee a clear path to profitability when your server bill is billions higher.

Best,

Jessica Lessin
Founder & Editor-in-Chief

Anthropic Lowers Gross Margin Projection as Revenue Skyrockets

Anthropic last month projected it would generate a 40% gross profit margin from selling AI to businesses and application developers in 2025, according to two people with knowledge of its financials. That margin was 10 percentage points lower than its earlier optimistic expectations, though it's still a big improvement from the year before.

The lower-than-expected gross profit margin resulted from the costs of running Anthropic models for paying customers, in a process known as inference, on servers from Google and Amazon. Those inference costs were 23% higher than the company had anticipated, the projections showed. Anthropic calculates gross margins by subtracting inference costs and other costs of selling its products.

A message from Resource Advisor+ Before AI Was a Buzzword, We Were Optimizing for Energy Schneider Electric has guided global energy and sustainability for decades. Now, Resource Advisor+ brings that expertise into an AI-native platform that unifies carbon accounting, supply chain visibility, risk management, and compliance. Built for leaders who demand both innovation and trust. See how it works.