First-Half Net Loss Grows, but Anthropic Plans 1 Quadrillion Won in Infrastructure Investment (Comprehensive) - The Asia Business Daily
WSJ Reports After Reviewing IPO Registration Papers
Anthropic Is Negotiating Infrastructure Deals Worth Up to $150 Billion
Plans to Spend Up to $750 Billion on Infrastructure
Anthropic, the company that makes the AI chatbot Claude, plans to invest up to $750 billion in infrastructure. Its net loss in the first half of this year was bigger than its loss for all of last year. This has raised new concerns about whether the company can afford such huge costs. However, news of its large investment plans has also raised hopes that cloud computing and chip companies could benefit.
Anthropic Plans to Invest Up to $750 Billion in Infrastructure
Anthropic CEO Dario Amodei speaks at an event in New Delhi, India, on Feb. 19, 2026 (local time). Reuters-Yonhap
View original imageAccording to a report published by The Wall Street Journal (WSJ) on October 20, local time, Anthropic's registration papers for its planned initial public offering (IPO) show that the company must pay $450 billion for cloud computing contracts over the next 7 to 10 years. The WSJ reported that many of these contracts cannot be ended early. Anthropic must pay for them even if it does not use all the services. This means the company could still owe hundreds of billions of dollars if demand for AI drops sharply or fewer people use its AI models.
Anthropic has signed contracts worth $111 billion with Google Cloud, $110 billion with Amazon Web Services (AWS), and $31 billion with Microsoft (MS). It has also agreed to pay $160 billion to lease Google's AI chips through Broadcom. When these deals and other cloud contracts are added together, the total value of its infrastructure agreements is more than $600 billion.
Anthropic is also negotiating more contracts to secure infrastructure worth up to $150 billion. If these deals are completed, its total spending commitments for infrastructure could reach $750 billion, the WSJ reported.
In its registration papers, Anthropic said it had asked cloud providers such as AWS, Google, and Microsoft for more computing power than they were willing to supply. These companies are expanding their own AI businesses and see Anthropic as a competitor. Anthropic said this was one reason it could not get more computing capacity.
International news outlets have raised concerns about whether Anthropic can afford these huge, long-term infrastructure contracts. The Financial Times (FT) said the financial pressure could increase if the company's revenue stops growing as quickly. Anthropic spent $9 billion on AI-related infrastructure in the first half of this year alone. That was four times the $2.3 billion it spent during the same period last year. It was also more than the $7.3 billion it spent during all of last year.
Net Loss Bigger Than Last Year's
As its expected future spending grows, Anthropic reported a net loss of $80.7 billion, or about 107 trillion won, in the first half of this year. However, $72 billion, equal to 89% of the total loss, came from accounting changes in the value of financial products used to raise money, including convertible bonds. This loss was almost twice the $42 billion loss reported for all of last year. More than $34 billion of last year's loss also came from changes in the recorded value of financial assets and liabilities.
If these changes in recorded value are left out, Anthropic's actual loss is estimated at about $8.7 billion. Even so, this amount is greater than the company's loss for all of last year. This has led to more questions about whether Anthropic can afford its huge spending plans.
Another risk is that Anthropic gets a large share of its income from only a few customers. The company said that about 20% of its revenue in the first half of this year came from two customers. It expects a "substantial portion" of its revenue to keep coming from a "limited number of customers" for some time. Revenue in the first half of this year rose 14.8 times, reaching $16.3 billion compared with $1.1 billion during the same period last year.
Anthropic also listed local opposition to building data centers as a business risk. Most of its data centers are in the United States, and it has signed some letters of intent to build facilities in other countries. The company said that opposition groups were becoming "more organized and coordinated across multiple regions." It also warned that its high profile "could attract particular attention to businesses associated with us."
The IPO prospectus also revealed how much the company's top executives were paid. Chief Executive Officer (CEO) Dario Amodei and his sister, President Daniela Amodei, each received a base salary of about $655,000 last year, listed as 869.19 million won. Including stock and stock options, Dario Amodei received about $18 million, or 23.886 billion won, while Daniela Amodei received about $16.5 million, or 21.8955 billion won. Their total pay was similar to the middle level of pay for CEOs at S&P 500 companies, but somewhat lower than the pay of executives at technology companies. The founders also promised to give 80% of their respective ownership stakes to charitable causes. On October 18, Anthropic announced that it would give 5% of its total shares to a newly created "public benefit fund." The fund can be used to make charitable donations or support "specific groups."
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.
