Where Does Anthropic’s Revenue Go? – To the Rising Costs of AI

By
Anthropic

Anthropic, the company behind the Claude AI assistant, is generating billions in revenue, but developing and operating advanced artificial intelligence remains enormously expensive.

Anthropic spent approximately $1.04 for every dollar of revenue generated during the first half of 2026. The company reported $16.3 billion in revenue but recorded an operating loss of around $650 million.

The figures reveal an important challenge facing the AI industry: even as demand for AI services grows rapidly, companies must invest heavily in computing infrastructure to keep up.

Computing Infrastructure Takes the Largest Share

Computing power is Anthropic’s biggest expense.

During the first six months of 2026, the company spent $8.9 billion on computing and infrastructure. Of that, $2.4 billion supported revenue-generating activities, while about $6.5 billion went toward model training, research, and free services.

Although Anthropic has become more efficient, reducing compute costs from $2.01 to $0.48 per revenue dollar over the past year, future expenses could increase significantly.

Morningstar estimates that Anthropic has approximately $819 billion in compute commitments, including $278 billion that could be canceled under certain conditions.

These agreements are intended to secure enough computing capacity to meet growing demand, but they also create financial risks if future revenue falls short of expectations.

The Companies Behind Anthropic’s Computing Power

One of Morningstar’s most interesting findings concerns where Anthropic’s money goes.

Much of its infrastructure spending flows to major technology companies, including Amazon, Google, Microsoft, and Broadcom, as well as specialized GPU cloud providers.

Many of these companies are also investors, lenders, or commercial partners of Anthropic.

This creates an interconnected financial relationship in which technology companies provide capital to Anthropic while simultaneously earning revenue from the computing infrastructure it purchases.

Cloud marketplaces also play an important role. Nearly half of Anthropic’s revenue comes through platforms such as Amazon Web Services, Microsoft Azure, and Google Cloud.

Morningstar estimates that these platforms collect approximately 15% of sales made through their marketplaces.

The arrangement benefits both sides as demand continues to grow. However, slower AI adoption could create financial pressure for Anthropic and its infrastructure partners.

High Revenue With a Relatively Small Workforce

Another notable finding is Anthropic’s revenue per employee.

The company employed approximately 5,300 people during the first half of 2026, with personnel expenses reaching $3.15 billion. About 70% of these expenses were stock-based compensation rather than cash salaries.

By September 2026, Anthropic’s trailing 12-month revenue per employee had reached $6.6 million, compared with $1.4 million a year earlier. These figures illustrate how AI companies can generate substantial revenue with relatively small teams, even while facing exceptionally high infrastructure expenses.

What This Means for Businesses Using AI

Anthropic’s financial structure raises questions about the long-term economics of AI services.

For businesses adopting AI, including e-commerce companies, computing costs are a key consideration. Applications such as product classification, catalog enrichment, customer service, and automated content creation must deliver enough value to justify their operating expenses.

The quality of the information AI systems use also matters. Structured product specifications, descriptions, and attributes, such as those provided by Icecat, can support more accurate and efficient AI-powered ecommerce workflows.

Rather than relying exclusively on increasingly powerful and expensive AI models, businesses may benefit from combining appropriate technology with reliable data and clearly defined use cases.

Can AI Revenue Keep Up With Spending?

Anthropic’s rapid growth demonstrates strong commercial demand for advanced AI, particularly among enterprise customers.

However, Morningstar’s analysis shows that the company’s future profitability depends heavily on maintaining revenue growth while managing substantial infrastructure commitments.

With more computing capacity scheduled to become available in 2027 and 2028, the financial pressure on AI developers could increase.

For the wider technology industry, the central question is whether growing demand for AI services can continue supporting the enormous investments required to develop and operate them.

manual thumbnail3

Manual for Channel Partners: Icecat Live HTML

Icecat Live is a (free) service that enables you to insert real-time produc...
 June 10, 2022

Icecat Add-Ons Overview. NEW: Claude AI, ChatGPT, AgenticFlow.AI, Mindpal.space and BoltAI

Icecat has a huge list of integration partners, making it easy for clients ...
 September 3, 2025
LIVE JS

How to Create a Button that Opens Video in a Modal Window

Recently, our Icecat Live JavaScript interface was updated with two new fun...
 November 3, 2021
New Standard video thumbnail

Autheos video acquisition completed

July 21, Icecat and Autheos jointly a...
 September 7, 2021
Manual How to Import Free Product Content Into Your Webshop via Icecat

Manual: How to Import Free Product Content Into Your E-commerce System via Icecat API

This guide is intended for developers working with Icecat via API. The docu...
 May 24, 2024