A note before the numbers: this is not financial advice and we are not experts on the IPO. Chasing Creative is not affiliated with Anthropic. Every figure below comes straight from Anthropic's preliminary prospectus (Form S-1/A) filed with the SEC on November 10, 2026. It is a preliminary document and the figures may change. Page numbers refer to the PDF of the filing.
Anthropic, the company behind Claude, has filed to go public. Most of the coverage is about the stock. We read it for a different reason: the filing is one of the clearest pictures yet of how much physical infrastructure the AI industry is committing to, and where that buildout is getting stuck.
If you design, build, power, cool, equip or operate data centers, these are the parts worth your time.
Watch the two minute version
Prefer YouTube? Watch the video there.
The scale
- About $900.9 billion in committed spend, substantially all of it for compute and infrastructure (p. 52).
- About $451.6 billion of that is expected before the end of 2030, with most of the rest between 2031 and 2045 (p. 52).
- About $70.7 billion comes due in the 12 months after September 30, 2026 (p. 52, 221).
- About $451.2 billion in leases signed but not yet started, generally beginning between 2026 and 2028, with an average term of about 10 years (p. 221).
- Compute and infrastructure cost $9.9 billion in the third quarter of 2026 alone. For all of 2025 it was $7.3 billion (p. 16, 20).
What it means: one AI company's order book is now a multi-decade construction program. Every dollar lands on a site somebody has to design, power and build.
Who is building it
- Google: at least about $121.3 billion through July 2033 (p. 222).
- Amazon: at least $110.0 billion through April 2036 (p. 222).
- Microsoft: at least $47.2 billion from November 2026 to June 2033 (p. 222).
- Broadcom: about $161.2 billion in equipment lease rent, excluding interest, for hardware deployed in dedicated data center facilities operated by others (p. 223).
- Newer providers: up to $120.5 billion with SpaceXAI, up to about $44.6 billion with Nscale and about $41.3 billion with Lambda (p. 222, 223).
What it means: the commitments run seven to ten years, which gives everyone who supplies those campuses long visibility. It also means newer names are being trusted with very large builds, and trust is a brand problem as much as an engineering one.
The bottlenecks the filing names
The risk factors read like a list of what our clients make and do.
- Electrical equipment. High-voltage transformers, switchgear, turbines, generators and substations face extended delivery times and price increases (p. 83).
- Long order windows. Some data center equipment comes from only a few manufacturers and must be ordered years in advance (p. 83).
- The grid. The filing says grid infrastructure was not designed for the rise in demand from AI, and that new generation, transmission and interconnection can take years (p. 83). Interconnection queues for large new loads have lengthened substantially in many regions (p. 84).
- People. Shortages of electricians, facility technicians and other skilled trades can stretch schedules and raise costs (p. 85).
- Community. Data center growth has drawn scrutiny over land, water and power, and some jurisdictions have paused new builds (p. 85, 86).
What it means: buyers are choosing suppliers years before a site opens. The firms that are easy to find, easy to understand and able to show proof of delivery are the ones that get the call.
The strategy shift
The filing describes a hybrid model: buying compute as a cloud service, and also placing its own hardware in dedicated facilities leased and operated on its behalf. Its first data center facility became operational this year (p. 256). It also notes that data centers take longer to deliver but are a smaller share of total compute cost, so capacity is secured early while larger chip purchases wait (p. 256).
What it means: buildings and power get committed first. Real estate, power and infrastructure brands sit at the front of the line.
What to do with this
None of this tells you what a stock will do, and we are not trying to. It tells you where an industry is putting its money and what is slowing it down. If your company is part of that supply chain, the demand is contracted and the buyers are looking now.
At Chasing Creative we help infrastructure brands get found and trusted with video, web and search. If you build, power or equip data centers and your story is not doing that work yet, let's talk.



