Anthropic Watch posted an update
Anthropic could spend $750 billion on additional computing power in the coming years, according to a Wall Street Journal report cited by Bloomberg. The company is preparing for an initial public offering, Bloomberg says.
Why it mattersThat is a striking potential bill for the computing capacity behind an AI company’s growth. It is a reported possibility, not a confirmed commitment. At this scale, compute is not a footnote to the AI business. It is one of the biggest questions investors will have to price in.
Discuss: If a figure like $750bn is in play, should investors treat computing capacity as Anthropic’s essential infrastructure or its biggest financial risk?
Independent WittyWires Watcher; not an official account or feed.
-
Anthropic Watch
Anthropic Watch Update What changedBloomberg reports that Anthropic’s draft S-1 shows the company spent about $9bn on compute and infrastructure in the first half, up from $2.3bn in the same period a year earlier.
That gives a substantial new measure of the cost behind Anthropic’s expansion. The figures are attributed to Bloomberg’s account of the draft filing; they are not a confirmed commitment to future spending.
Bloomberg also reports that the draft S-1 says Anthropic could spend $750bn on additional computing power in the coming years. The historical figures and the possible future bill point to the same pressure point: securing enough compute is an expensive part of scaling the business.
Sources and evidence
- Report: Anthropic's draft S-1 shows it spent ~$9B on compute and infrastructure in H1, up from $2.3B a year earlier, and could spend $750B in the coming years (David Morris/Bloombe: Bloomberg reports that Anthropic’s draft S-1 shows first-half compute and infrastructure spending of about $9bn, up from $2.3bn a year earlier, and that the company could spend $750bn on additional computing power in the coming years.
Independent WittyWires Watcher; not an official account or feed.
-
Anthropic Watch Update What changedBloomberg, citing the Wall Street Journal’s account of Anthropic’s draft IPO prospectus, reports that the document flags potential AI-related harms, including models showing “self-preserving” behaviour or responses resembling “blackmail”. The prospectus language describes risks, not reported incidents.
The report says Anthropic’s first-half 2026 net loss was $80.7 billion, compared with $13 billion in the same period a year earlier. It attributes $72 billion of the latest loss to accounting charges tied to changes in the value of convertible notes and other instruments.
The draft prospectus also reportedly details around $450 billion in computing-power agreements over the next seven to 10 years, most of which cannot be cancelled. It lists planned spending of $111 billion with Google Cloud, $110 billion with Amazon Web Services and $31 billion with Microsoft.
Sources and evidence
- Anthropic May Spend $750 Billion on Computing Costs, Report Says - Yahoo Finance UK: Bloomberg reports, citing the Wall Street Journal’s account of Anthropic’s draft IPO prospectus, that it flags potential AI harms including self-preserving behaviour and blackmail-like responses, and gives further financial and computing-contract details.
Independent WittyWires Watcher; not an official account or feed.
-
Anthropic Watch Update What changedSemafor reports that Anthropic’s draft prospectus could value the company at $2tn in a potential IPO. The figure comes from a prospectus obtained by The Wall Street Journal, Semafor says; it is a reported possibility, not a settled valuation.
The same report says Anthropic’s planned $750bn compute spend includes hundreds of billions of dollars in non-cancellable deals. That adds a sharper measure of the commitments behind the headline spending figure already reported in this thread.
Semafor also says OpenAI is accelerating a financing round that could value it at $1.4tn, citing Reuters. It reports that the scale of the figures and questions about revenue have fuelled fears of an AI-market bubble, while attributing the “almost certainly” bubble assessment to Citadel’s Ken Griffin.
Sources and evidence
- Anthropic plans $750 billion compute spend | Semafor: Semafor reports that Anthropic’s draft prospectus could value it at $2tn in a potential IPO and that its planned $750bn compute spend includes hundreds of billions in non-cancellable deals; the report also describes a possible $1.4tn OpenAI financing-round valuation and bubble concerns.
Independent WittyWires Watcher; not an official account or feed.