Anthropic has revealed nearly $900 billion in long-term spending commitments for computing power and infrastructure in an updated filing ahead of its planned initial public offering.
Anthropic has disclosed approximately $895.4 billion in aggregate purchase commitments and lease agreements for computing power and infrastructure, according to an updated S1 filing ahead of a planned initial public offering this month.
In the three months ending September 30, Anthropic reported revenue of approximately $17 billion, which represents a 47 percent increase from the $11.6 billion reported in the previous quarter. This marks a significant increase from the same period a year ago, though the company noted that revenue growth has slowed sequentially compared to earlier in the year.
Anthropic's spending on chips, power, and infrastructure rose 77 percent, increasing from $5.6 billion to $9.9 billion. Data center expenses reached 58 percent of revenue in the third quarter, up from 48 percent in the second quarter. The company reported a net loss of $23.3 billion for the July-September quarter, which it stated was driven primarily by non-cash accounting charges linked to its rising valuation. The company's net loss shrank from $66.2 billion in the second quarter.
According to the filing, the $895.4 billion in commitments is largely related to compute and infrastructure. The breakdown of these obligations includes:
Specific components of these commitments include $321 billion in leases that have not yet started and $265.6 billion in purchase agreements. Anthropic has also committed to spending more than $3 billion a month with SpaceX alone.
Anthropic is preparing for an initial public offering (IPO) that backers anticipate could value the company between $1.5 trillion and $2 trillion. The company plans to raise up to 552.9 billion yen ($3.5 billion) through a share sale to Japanese investors to support working capital and capital expenditure.
In its updated risk factors, Anthropic warned that its latest AI models "may have already advanced" at a rate that triggers the implementation of new internal safety measures. The company stated its future models "may imminently meet our Responsible Scaling Policy’s definition of ‘dramatically accelerated progress attributable to AI’". Anthropic also noted that it has in the past not always identified significant accelerations in its rate of development until after they occurred.
Sources
This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.
The day's checked news, by email
One short email a day with the top stories. Free, and you can unsubscribe in one click.
Or follow for the latest news:
No comments yet. Start the discussion.
Former Secretary of the Army Dan Driscoll joins OpenAI to advise on the responsible use of advanced AI for national and cyber security alongside U.S. allies.
OpenAI has introduced Pando, a new Git-compatible backend designed to handle high-concurrency writes and maximize read throughput for its monorepo.
