The commitments disclosed in the filing amount to multi-year revenue visibility for Anthropic’s infrastructure suppliers, with Google, Amazon Web Services, Microsoft and Broadcom each tied to contracts running into the tens or hundreds of billions of dollars, much of it noncancelable. That may support the case that AI infrastructure demand is contracted rather than speculative, though it also concentrates counterparty risk on a single loss-making buyer. For prospective IPO investors, the debate is likely to centre on whether revenue growth can keep outrunning compute costs, and on customer concentration, with two clients contributing around a fifth of first-half sales. A valuation of up to $2 trillion would leave little room for a slowdown on either front.
Anthropic’s draft prospectus shows a company growing faster than its costs, but committed to an infrastructure bill so large that its suppliers, customers and regulators all become part of the investment case.
Summary:
First-half 2026 revenue of about $16.3 billion, up from about $1.1 billion a year earlier, according to a draft prospectus viewed by the Wall Street Journal (gated)
Net loss of about $81 billion for the half, of which roughly $72 billion came from accounting charges tied to convertible notes and other financing instruments; the prior-year half-year loss was about $13 billion
Compute and infrastructure spending of nearly $9 billion in the half, more than in all of 2025
Infrastructure commitments exceed $600 billion, with further deals under negotiation that could add up to $150 billion
Around a fifth of first-half revenue came from two customers
Risk factors include AI safety and liability, possible uninsurable catastrophic incidents, constrained compute access and organised opposition to data centres
Anthropic grew revenue nearly fifteenfold in the first half of 2026 while committing to more than $600 billion in future computing infrastructure, according to a draft initial public offering prospectus reviewed by the Wall Street Journal. The company declined to comment on the report.
The artificial intelligence developer generated about $16.3 billion in revenue in the six months to 30 June, compared with about $1.1 billion a year earlier, the Journal reported. That is already well above the roughly $4.6 billion it brought in during all of 2025. Spending on computing capacity and infrastructure rose to nearly $9 billion in the half, from about $2.3 billion a year earlier, and exceeded the company’s full-year 2025 outlay of around $7.3 billion.
The headline loss is far larger. Anthropic recorded a net loss of about $81 billion for the half, against roughly $13 billion a year earlier and about $42 billion for the whole of 2025. Around $72 billion of that reflected accounting charges linked to changes in the value of convertible notes and other instruments used to raise funding, rather than operating costs.
The scale of Anthropic’s infrastructure commitments is the filing’s most striking feature. According to the Journal, the company owes around $450 billion under cloud agreements over the next seven to 10 years, most of which cannot be cancelled. That includes about $111 billion with Google, $110 billion with Amazon Web Services and $31 billion with Microsoft, plus up to $86 billion for Nvidia chip capacity from Nscale and Lambda. A separate obligation of more than $160 billion to lease Google chips through Broadcom lifts the total above $600 billion, and deals still under negotiation could add up to $150 billion more.
Even so, the company said its major cloud partners had not provided as much capacity as it sought, and that those partners increasingly view it as a competitor. Anthropic is supplementing that capacity with leased data centres and third-party operators, some of which it said have short track records.
Revenue is also concentrated. Two customers accounted for about a fifth of first-half sales, and the company expects a large share of revenue to keep coming from a small group of clients.
Risk factors make up roughly a third of the document, the Journal reported. They include the possibility of misaligned AI models causing harm, liability for damages such as data theft, and incidents severe enough that insurance may not cover them. Anthropic also cited increasingly coordinated political opposition to data centre construction, noting that its public profile could draw attention to projects linked to it.
The filing comes as Anthropic prepares for a listing that could value it at up to $2 trillion, putting its growth rate, cost base and customer mix under far closer public scrutiny than before.
This article was written by Eamonn Sheridan at investinglive.com.
