By Anurag Gurtu
When a company delays its IPO to show investors better numbers, that’s a strategy. When the NYC pension fund publicly asks you to delay it over AI safety concerns, your chief rival’s CEO calls his own IPO “ill-advised” and postpones it until at least next year, tech stocks are sliding on open-weight model fears, and your own prospectus warns of “existential risks to humanity” from the product you’re selling, that’s not a strategy.
That’s a market sending you a message.
Anthropic‘s numbers are extraordinary. Its third-quarter 2026 revenue was reportedly $16.8 billion — up 45% quarter on quarter. That means the company’s annualized revenue run rate is roughly $67 billion.
What the S-1 actually says
Here are the numbers from Anthropic’s leaked prospectus:
- $4.6 billion: Anthropic’s full-year 2025 revenue.
- $7.33 billion: What Anthropic spent on compute and infrastructure in 2025 alone — more than its total revenue.
- $42 billion: Anthropic’s net loss in 2025. The roadshow will immediately explain that $34 billion of it was a noncash accounting charge — an adjustment to the estimated value of financing instruments that could eventually convert into shares.
- $518 billion: How much Anthropic expects to spend, at least, over the course of a decade on AI infrastructure with six partners, including Google, Microsoft and Amazon, according to the confidential IPO prospectus reviewed by Reuters. About 80% of the total is noncancelable or payable regardless of usage.
- $2 trillion: That’s the IPO valuation Anthropic is still targeting for November, a figure representing 435x its 2025 revenue.
One more number from the S-1 that the roadshow will bury: A quarter of Anthropic’s 2025 revenue came from two unnamed customers.
Three signals the market is sending
The IPO delay is the headline. It isn’t the only signal worth reading.
Signal one: NYC Comptroller Mark Levine. Last week, the supervisor of New York City’s pension funds — which reportedly hold roughly $7 million in Anthropic exposure today, but could hold dramatically more if Anthropic enters the Russell 3000 post-IPO — sent a letter to Anthropic CEO Dario Amodei urging the company to pause the IPO entirely.
The risks he cited are the same ones Anthropic’s own prospectus flags: increasingly autonomous AI systems that could behave in unexpected ways, create security problems, and in the document’s own language, pose “existential risks to humanity.”
Signal two: Sam Altman delays OpenAI’s IPO. OpenAI has said it will not go public in 2026. CEO Altman called an IPO at this moment “ill-advised” given AI safety concerns, and while that framing is partly strategic — positioning OpenAI as the more responsible actor — it’s also a real signal when the CEO of the second-most valuable AI company in the world looks at current conditions and decides to wait.
Signal three: The chip selloff. Last week’s $2 trillion chip-stock selloff was driven partly by concern over AI-buildout costs. Gartner’s forecast helps explain the concern: Through 2030, it sees at most $3.6 trillion in revenue attributable to AI data centers, versus $8.5 trillion needed to cover planned investment, financing costs and a 12% return.
The open-weight problem
Here is the structural argument that no amount of Q3 revenue growth resolves: Anthropic’s $2 trillion valuation requires closed-model AI to maintain a durable pricing premium over open-weight alternatives. That premium is being systematically dismantled.
Open-weight AI is challenging the economics behind closed-model valuations. Enterprise adoption rose from 23% to 67% in 2026, with companies reporting 70%–90% cost savings. DeepSeek R1 offers frontier-level performance on most enterprise benchmarks at 7 cents per million input tokens; independent analysis puts Anthropic’s comparable input and output pricing at roughly 18x and 23x higher, respectively.
DeepSeek’s weights are free to download, allowing enterprises to self-host without a model-provider API contract or per-token billing, while the capability gap with closed models reportedly closes within six months of a frontier release.
Against that backdrop, New Constructs — the research firm that called Anthropic’s proposed offering “an unprecedented test of investor gullibility” — calculated that justifying a $2 trillion valuation would require an immediate 10% after-tax operating margin sustained for a decade, extraordinary revenue growth, and $358 billion in annual after-tax operating profit by 2035.
Why this is worse than 1999
The dot-com bubble had a genuine excuse. Nobody knew what the internet would become. The AI IPO wave of 2026 does not have that excuse.
Worldwide spending on AI is forecast to reach $2.59 trillion in 2026, up 47% year over year, according to Gartner. Yet a July 2025 MIT NANDA report estimated that 95% of the organizations it examined were getting no return from generative AI. Meanwhile, KPMG said in its own report earlier this year that just 7% of surveyed leaders say they’ve established ROI from AI.
This is the context in which Anthropic is asking investors to pay 435x 2025 revenue for a company with $518 billion in locked infrastructure commitments, 25% revenue concentration in two unnamed clients, and a pricing moat that open-weight models are closing in six-month cycles.
The delay from October to November doesn’t change any of those numbers. It adds one quarter of revenue data.
What actually survives this correction
Here is what no AI IPO roadshow will say out loud: The biggest enterprise AI winners will not necessarily own the most expensive model. They will own the architecture that turns any model into governed, auditable, measurable outcomes — through proprietary enterprise data, domain-specific fine-tuning, multiagent orchestration and controls that make autonomous execution trustworthy.
That value does not depend on keeping model capabilities scarce or securing a $2 trillion valuation. It compounds in production, where customers understand that the model is not the product. The deployment is.
The dot-com crash did not happen because the internet was bad. It happened because the price was wrong. AI investors face the same distinction: Real technology does not automatically justify its valuation. A strong quarter and a one-month delay do not close that gap. November simply gives Anthropic more time to explain the math.
Anurag Gurtu, the co-founder and CEO of Airrived, is an entrepreneur and product leader with more than 20 years of experience across cybersecurity and AI. He has built three generative AI companies, including one that reached unicorn status. Gurtu is known for turning advanced AI into scalable, real-world platforms, and previously contributed to successful exits, including acquisitions by Splunk, Tripwire and Sun Microsystems, and a public company, FireEye. He holds a master’s degree from the University of Southern California and CISSP certification.
Related Crunchbase queries:
Illustration: Dom Guzman