Anthropic at $1.4 trillion? My Valuation Is Under $1 billion
Anthropic targeted an IPO valuation of $1.4 trillion while committing $900.9 billion to infrastructure, prompting skepticism over required earnings.
By Dillip Chowdary • Nov 19, 2026 • Source: ludovicphalippou.substack.com

At a glance
- Target valuation
- Up to $1.4 trillion fully diluted
- Prospectus filing
- November 10, 2026
- Compute commitments
- $900.9 billion aggregate
- Near-term compute
- $451.6 billion through 2030
- Phalippou valuation
- Under $1 billion
- Inference benchmark
- $0.30 per million tokens (Revolut)
In this post
Anthropic filed registration documents on November 10 targeting an initial public offering valuation of up to $1.4 trillion on a fully diluted basis while disclosing hundreds of billions in compute obligations. Financial author Ludovic Phalippou challenged that target in ludovicphalippou.substack.com's report, arguing the company's realistic valuation is under $1 billion when weighed against infrastructure demands and enterprise switching. The analysis contrasts Wall Street forecasts with physical capital limits facing frontier model builders.
This report details the financial metrics, electrical grid bottlenecks, and corporate migration trends shaping Anthropic's public filing for enterprise buyers and market analysts. Readers will evaluate the revenue scale needed to sustain trillion-dollar capitalizations, the power constraints stalling data centers, and real-world results from engineering teams deploying open-weight models.
01Public filings and revenue math
Public market analysts attempt to justify Anthropic's $1.4 trillion valuation target by discounting projected cash flows through 2045 with perpetual growth rates between 3% and 3.5%. Phalippou noted in his book, The Science of Valuation in Private Markets, that while private equity enforces disciplined entry prices, long-range public market models rely on speculative input assumptions. Discounting cash flows decades ahead for a product that barely existed a few years ago produces unreliable valuations.
Supporting a $1.4 trillion equity valuation without assuming perpetual growth requires substantial earnings comparable to mature industrial giants. At an illustrative multiple of 25 times annual earnings, Anthropic would need to generate approximately $56 billion in sustainable annual net profits. Assuming an optimistic 20% net margin, that requires $280 billion in annual revenue, an outcome unsupported by current evidence.
02Data center power and capital commitments
While promoted as a scalable software revolution, frontier artificial intelligence increasingly resembles a capital-heavy infrastructure program. Training and serving modern models demands massive data centers, specialized processors, cooling systems, and scarce electrical power. The International Energy Agency estimated data centers consumed 415 TWh of electricity in 2024 and projected consumption will reach 945 TWh by 2030, exceeding Japan's current annual electricity usage.
Electricity grids cannot easily supply planned computing clusters, and local opposition over utility bills and environmental impacts has intensified. Research reported by Bloomberg revealed that local opposition blocked or delayed 45 United States data center projects valued at approximately $68 billion in the second quarter of 2026 alone. Transmission lines require years to build, creating severe geographic constraints for expanding facilities.
Anthropic's November 10 prospectus revealed $900.9 billion in aggregate compute and infrastructure commitments, including agreements entered into after September 30. Approximately $451.6 billion must be spent by the end of 2030, with additional obligations extending through 2045. These multi-year commitments require continuous cash generation to service computing hardware that quickly becomes obsolete.
03Leaderboard point spreads and memory export
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
Developing frontier models requires enormous capital and technical expertise, creating significant barriers to initial entry. However, these barriers do not guarantee pricing power once multiple well-capitalized competitors enter the market. Microsoft, Google, OpenAI, xAI, Meta, and several Chinese developers are investing aggressively in proprietary and open-weight systems, rapidly eroding individual technological advantages.
Stanford's 2026 AI Index reported that in March 2026, the four leading model providers were separated by only 22 points on the Arena leaderboard. Anthropic scored 1,503, followed closely by rivals with near-identical benchmark performances that demonstrate how quickly competitive leads rotate across the sector.
| Provider | Model Developer | March 2026 Arena Score |
|---|---|---|
| Anthropic | Anthropic | 1,503 |
| xAI | xAI | 1,495 |
| 1,494 | ||
| OpenAI | OpenAI | 1,481 |
Switching costs are also lowered by emerging data portability features across commercial tools. Claude now allows users to export saved memories and transfer contextual profiles directly to competing providers. While deep enterprise workflow integrations carry migration overhead, portable user memory reduces vendor lock-in across day-to-day operations.
04Engineering costs and open-weight migration
Enterprise software teams facing high commercial API costs have started evaluating self-hosted alternatives for internal workflows. Financial technology company Revolut, which employs approximately 2,000 software engineers, previously utilized commercial coding tools including Anthropic's Claude Code and OpenAI's Codex. However, substantial monthly inference expenses prompted the company to reassess its vendor commitments.
In October 2026, David Abitbol, head of AI operations at Revolut, announced the company shifted to GLM-5.3, an open-weight model from Chinese firm Z.ai. Running on Revolut's controlled infrastructure, the deployment reduced the company's AI coding inference bill by roughly 90%, achieving a blended cost around $0.30 per million tokens.
Note: Revolut noted these figures represent internal operational metrics that do not necessarily incorporate all hosting and staffing overhead on an equivalent basis. The institution has not abandoned commercial systems entirely, continuing to run proprietary frontier models for select programming tasks alongside its self-hosted stack.
05Hardware supplier profits and long-term margins
The continuous capital demands of frontier artificial intelligence raise basic questions about who captures economic value across the supply chain. Chip manufacturers, data center operators, and utility providers command significant pricing power over model builders. If remaining competitive requires a permanent investment race, model developers may experience margin profiles far below traditional software businesses.
Frontier models face rapid hardware obsolescence and constant competitive pressure, preventing capital expenditures from tapering off into low-cost maintenance. Without defensible assets that cannot be replicated by well-funded competitors or open-weight alternatives, sustaining premium profit margins over decades remains difficult. Anthropic's multi-decade compute liabilities highlight the ongoing financial risks facing pure-play artificial intelligence developers.
Frequently asked questions
What valuation is Anthropic targeting for its initial public offering?
According to its November 10 filing, Anthropic is targeting an IPO valuation of as much as $1.4 trillion on a fully diluted basis.
Why does Ludovic Phalippou value Anthropic at under $1 billion?
Phalippou argues that sustaining a $1.4 trillion valuation requires $56 billion in net profits and $280 billion in annual revenue, while Anthropic faces $900.9 billion in compute commitments, narrow benchmark moats, and rapid open-weight enterprise migration.
What are Anthropic's reported infrastructure and compute commitments?
Anthropic's prospectus disclosed approximately $900.9 billion in total infrastructure commitments, with $451.6 billion scheduled to be spent by 2030 and substantial obligations continuing through 2045.
How did Revolut reduce its AI coding costs?
Revolut replaced its default coding tools with the open-weight GLM-5.3 model on self-hosted infrastructure, reducing its AI coding inference bill by roughly 90% to about $0.30 per million tokens.
Sources
Author
Dillip Chowdary
Writes Tech Bytes coverage of AI, engineering, and the tools that actually ship. Editor of Tech Pulse Daily.
Related on Tech Bytes
Advertisement