Anthropic's confidential draft IPO prospectus, leaked to Reuters on September 28, commits the artificial intelligence lab to spending at least $518 billion on cloud services and data centers over the next decade, with about 80 percent of those obligations binding and non-cancelable, Reuters reported after reviewing the filing. The structure concentrates spending with four infrastructure partners, with $161.2 billion allocated to Broadcom, $111.1 billion to Google, $110 billion to Amazon, and $31.4 billion to Microsoft, the filing shows. For Anthropic, whose 2025 revenue reached $4.6 billion against a $42 billion net loss, the commitments effectively lock the company into a decade of payroll for chip engineers, data center construction crews, and the long tail of trades that follow AI infrastructure money.

Total AI infrastructure commitment
$518 billion over a decade, per the leaked draft S-1
Share that is binding and non-cancelable
About 80 percent
Broadcom chip-related lease commitment
$161.2 billion
Google cloud commitment
$111.1 billion
Amazon cloud commitment
$110 billion
Microsoft cloud commitment
$31.4 billion
Anthropic 2025 revenue
$4.6 billion
Anthropic 2025 net loss
$42 billion
Q2 2026 revenue
$11.5 billion, more than double the $4.73 billion reported in Q1 2026
Customer concentration
25 percent of revenue from just two clients
Open Anthropic roles
More than 640 across the company's careers page in September 2026

A decade of cloud and chip commitments

The draft prospectus, which Anthropic filed confidentially with the SEC on June 1 and which has not yet appeared on the public EDGAR database, lays out the company's spending through 2035 across six infrastructure partners. Reuters' review of the document, summarized by Yahoo Finance and Forkast News on September 29, showed that about 80 percent of the $518 billion is binding, meaning Anthropic must pay regardless of how the business performs. The $161.2 billion committed to Broadcom covers chip-related equipment leases, while the larger cloud deals with Google, Amazon, and Microsoft cover the compute and hosting that powers the Claude model family. Anthropic reported $20.28 billion in cash and cash equivalents at the end of last year, well short of the obligations it has signed up for.

What it means for hiring across the AI supply chain

For workers, the scale of the deal sets the floor on a decade of construction and operating hiring tied to AI infrastructure. Anthropic itself is hiring aggressively to deliver on the S-1 roadmap: its careers page listed more than 640 open roles in September 2026, including a recently advertised Staff Software Engineer position on the GTM AI Engineering team that pays $320,000 to $405,000 a year. The bigger hiring effect sits upstream. Data center operators like Google, Amazon, and Microsoft will need to staff and build out the capacity Anthropic has pre-committed to, which means electricians, HVAC technicians, network engineers, security staff, and the construction trades that build the buildings in the first place. Reuters' breakdown of the prospectus lists roughly $111.1 billion to Google, $110 billion to Amazon, and $31.4 billion to Microsoft, plus the $161.2 billion Broadcom figure, a hiring pipeline that runs well beyond Anthropic's own payroll.

The financial pressure behind the commitment

The scale of the obligations sits in tension with Anthropic's current financials. The company posted $4.6 billion in revenue in 2025 and a $42 billion net loss, according to the prospectus reviewed by Fortune on September 29. Quarter-over-quarter revenue grew sharply, with $11.5 billion in the second quarter of 2026 versus $4.73 billion in the first, but the prospectus also flags that 25 percent of revenue comes from just two customers, a concentration that creates risk if either client reduces spend. CEO Dario Amodei devoted more than a third of the S-1 to risk disclosures, including warnings about what the filing describes as existential risks from more advanced AI models, per Reuters. The combination of a $518 billion infrastructure obligation, a $42 billion annual loss, and customer concentration will shape whether Anthropic can keep staffing at the pace its S-1 assumes through the rest of the decade.

Primary source

Check the original source

Reuters exclusive on Anthropic $518 billion AI buildout is the source to consult for the underlying data, statement, ruling or live context.

Open Reuters exclusive on Anthropic $518 billion AI buildout

Sources and editorial note

This original Hidden Jobs analysis uses the report from Fortune coverage of leaked Anthropic IPO prospectus (published September 29, 2026) as a secondary source and points readers to the primary source for verification. Hidden Jobs is not affiliated with the organisations or sources mentioned in this story, and reported conditions, figures and policies can change.

More from the newsroom

Labour market AMD buys Fei-Fei Li's World Labs for $8.2 billion to build physical AI Labour market US Job Postings Turn Positive for the First Time in Nearly Four Years Labour market Apple Cuts Fitness+ Jobs Again in a Second Layoff Round
← Back to Hidden Jobs News Explore remote tech jobs