Anthropic’s IPO trillion-dollar question: can AI become a profitable business?
Key takeaways
- Anthropic gives investors direct exposure to AI models as Claude moves well beyond chat.
- Business adoption is soaring, but computing costs and competition threaten profitability.
- At roughly 1.5 trillion USD, the IPO price assumes remarkable future success.
For much of the AI boom, investors have bought the shovels: Nvidia's chips, cloud servers and data centres. Now they can invest in the intelligence those machines produce.
Anthropic, the company behind Claude, priced its initial public offering (IPO) at 83 USD per share on 18 November 2026, raising about 81 billion USD and surpassing SpaceX's 75 billion USD IPO in June 2026 to become the largest IPO in history.
The question is simple: can the intelligence pay for itself?
Claude wants to become indispensable
Claude writes software, analyses documents and helps businesses automate work. The opportunity is not simply another chatbot subscription. It is becoming part of how companies operate.
Demand is substantial. By September 2026, Anthropic had more than 11,900 customers generating at least 100,000 USD each over the previous year, up from more than 950 a year earlier. Revenue reached an annualised pace of 81 billion USD in October.
As Claude becomes embedded in companies' systems, changing providers may become more difficult. Customers could also spend more as they find new uses. Yet today's leading model can be overtaken tomorrow. OpenAI and Google are formidable rivals, while cheaper alternatives threaten pricing.
Anthropic's strongest competitive advantage may ultimately be its place inside customers' businesses, rather than its place on a leaderboard.
The costly business of intelligence
Anthropic's growth is extraordinary. So are its bills.
Third-quarter revenue approached 17 billion USD, but the company reported an operating loss of 567 million USD. On an adjusted basis, however, operating income reached 2.5 billion USD, suggesting improving underlying profitability once certain costs are excluded.
Computing costs remain the biggest challenge. Anthropic spent 18.8 billion USD on computing and infrastructure in the first nine months of 2026, while competition could push AI prices lower even as demand grows.
The real question is whether Anthropic can make intelligence cheaper to produce faster than it becomes cheaper to sell.
Beyond the opening bell
For investors, three developments deserve particular attention:
- Customer loyalty: are existing customers expanding their use of Claude, rather than simply trying it?
- Profitability: are computing costs falling relative to revenue, allowing operating margins to improve?
- Competition: can Anthropic retain customers and pricing power as rival models become cheaper and more capable?
Investors must also consider governance. Anthropic's founders and a separate benefit trust control key voting and board decisions, limiting ordinary shareholders' influence.
The shovels were only the beginning
Anthropic offers a direct stake in a leading AI developer. Its rapid growth shows that businesses are willing to pay for intelligence. But a successful product, a successful company and a successful investment are three different things.
Investors are already paying for Claude to become deeply embedded in corporate life, for computing to become more efficient and for competition to leave room for healthy profits. All are possible, but none is guaranteed.
The excitement of opening-day trading may reveal more about enthusiasm for AI than long-term value.
The boom began with companies selling the shovels. Anthropic's debut brings the harder question into focus: how much profit will the intelligence itself produce?