Anthropic's proposed Nasdaq listing has put an unusually large set of AI infrastructure commitments into public view, sharpening the market's focus on the companies supplying data centers, chips, memory, power equipment and cooling systems. In the three trading sessions after the company filed its S-1 registration statement with the U.S. Securities and Exchange Commission on Oct. 26, Vertiv shares rose 25.0%, CoreWeave gained 17.9%, Oracle advanced 14.4%, Micron added 12.0%, Broadcom climbed 9.4% and Nvidia rose 2.1%, according to Nasdaq market data.
The S-1 describes eight named computing arrangements totaling roughly $657 billion, alongside additional TPU capacity expected to cost between $125 billion and $150 billion once priced. Those figures help explain media estimates of approximately $807 billion in total capacity commitments, though Anthropic did not provide one consolidated headline number in the filing.
At least $265.6 billion of the $657 billion group consists of fixed or minimum guaranteed obligations, with some contracts running for as long as 20 years. Based on Anthropic's reported trailing 12-month revenue of $19.708 billion, those minimum commitments equal roughly 13.5 times annual revenue. The larger $657 billion figure is about 33 times that revenue.
That scale places the company's planned expansion among the biggest demand signals yet for the physical systems behind advanced AI. It also gives suppliers a clearer basis for booking future capacity, ordering components and arranging project financing, even where the timing of actual computing usage remains tied to Anthropic's growth.
Rapid revenue growth meets long-term infrastructure obligations
Anthropic's filing showed a business expanding at a pace rarely seen in enterprise software. Revenue increased from $376 million in the first quarter of 2025 to $11.555 billion in the second quarter of 2026, according to the S-1. The company reported 144.2% quarter-on-quarter growth and a 1,368% year-on-year increase for the latest quarter.
The revenue mix points to an AI market driven largely by metered consumption rather than conventional software contracts. During the first half, consumer-style token billing generated $13.1 billion, or 80% of revenue, while subscriptions accounted for $3.2 billion, or 20%. Anthropic reported that 72% of revenue came from the United States and 28% from international markets.
More than 80% of revenue was linked to models launched during the preceding six months. That concentration gives Anthropic a powerful growth engine when new models gain traction, but it also ties a large part of its revenue base to a fast product cycle in which customers can shift spending as competing models improve.
The company reported a net revenue retention rate of about 900%, meaning the same customer group was spending about nine times more after one year. Customers paying at least $100,000 annually increased from roughly 950 to more than 11,900. More than 1,900 customers were spending at least $1 million, while about 250 spent at least $10 million.
Customer concentration has eased as the commercial base expanded. The top five customers represented 36% of revenue in the first quarter, 25% over the first three quarters and 21% in the third quarter alone, the filing said. No customer accounted for more than 10% of third-quarter revenue, although earlier periods included two 2025 customers contributing 12% each and one first-half 2026 customer contributing 11%.
Costs are rising, but utilization is improving
Compute and infrastructure expenses rose 300% over the period covered in the filing, reflecting the cost of training and serving increasingly capable models. Anthropic reported second-quarter GAAP operating profit of $974 million, equivalent to an 8% operating margin, although the company recorded a $693 million operating loss for the first half.
First-half expenses totaled $16.979 billion, or 104% of revenue. Compute represented $8.945 billion, or 53% of total expenses; personnel was $3.152 billion, or 19%; sales, marketing and channels also represented 19% at $3.168 billion; and other costs totaled $1.714 billion. Research and development spending reached $9.741 billion, about 60% of revenue.
The filing's most constructive efficiency measure was revenue generated for each dollar of compute spending. That figure rose from $0.40 in the first quarter of 2025 to $2.08 in the second quarter of 2026, a 5.2-fold increase across six quarters. Anthropic also said revenue over the past 12 months increased about 1,200%, while headcount rose about 170%, lifting annualized revenue per employee from $1.4 million to more than $6.6 million.
These figures suggest the company has become more effective at monetizing computing capacity, but they do not remove the execution pressure created by multiyear obligations. The company had $10.8 billion in remaining performance obligations as of June 30, with about 75% expected to be recognized within 12 months.
Suppliers gain visibility into data-center demand
Vertiv's results gave the clearest immediate read-through from Anthropic's filing and the wider AI buildout. The power and cooling supplier reported Oct. 28 revenue of $4.313 billion, up 61% year on year, adjusted earnings per share of $2.21 and an adjusted operating margin of 25.2%. Organic orders rose more than 100%, and Vertiv raised the midpoint of its 2026 sales outlook by $650 million to $14.65 billion. Its shares gained 16.48% that day, according to Nasdaq data.
Power availability is becoming as consequential as chip availability for large-scale AI projects. GE Vernova reported $29.3 billion in orders, up 96%, and a backlog that rose $20 billion sequentially to $197 billion. Its gas-equipment reservation agreements increased from 116 gigawatts to 122 gigawatts, with at least 130 gigawatts projected by the end of 2026.
Eaton reported second-quarter electrical-segment backlog up 43% year on year and said Americas data-center revenue increased about 65%. Those figures were disclosed at the segment level and were not attributed to any specific Anthropic contract, but they show how data-center construction is reaching electrical distribution equipment beyond servers and accelerators.
Financing structure becomes a market checkpoint
Anthropic's plans also connect with a more complex financing system developing around AI capacity. Broadcom disclosed a five-year master lease with approximately $161.2 billion in rent payments, excluding interest, and a separate $63 billion special-purpose vehicle financing arrangement involving Bank of America, Blackstone, Citi and Morgan Stanley.
The SPV is intended to purchase XPU accelerators and network equipment for lease-back and is expected to issue about $21 billion of bonds. Broadcom also disclosed up to $4.2 billion in lease-financing capacity, carrying a standard 9% payment-in-kind rate and a 13% rate under specified revenue conditions. Outstanding principal was zero at the end of June.
Broadcom's operating results remain strong: fiscal third-quarter revenue reached $29.6 billion, up 85.5%, while AI semiconductor revenue increased 221% and free cash flow totaled $13.665 billion. Its Dec. 9 earnings report and expected SPV bond pricing are likely to offer a closer view of how readily capital markets will fund the infrastructure behind large AI commitments.
The Anthropic filing supports a more specific conclusion than claims that AI-related digital assets will automatically benefit from hardware spending. Its disclosed commitments are directed overwhelmingly toward conventional cloud capacity, semiconductors, networking, power and cooling. The public-market response has so far followed that contractual chain, rewarding suppliers with visible orders, backlog growth and financing arrangements rather than token categories without direct exposure to those projects.
Want deeper context on AI-driven markets and trading strategies? Explore our latest insights in the AI market analysis hub.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
