Price for the future

Anthropic's IPO will be a test case

The reported details of Anthropic’s initial public offering (IPO) highlight the stratospheric valuations investors are assigning to artificial intelligence (AI). The IPO documents also flag concern about AI’s tendency to go rogue. Anthropic, which declared $42 billion in 2025 losses, is hoping to raise funds at a valuation of $2 trillion. It is said to have devoted 80 pages of offer documents to grim warnings about the existential threat posed by AI. The prospectus warns that AI models could exhibit “self-preserving behaviours”, including attempts to “resist shutdown”, to “conceal or manipulate information” and behaviour “resembling blackmail”. The company also says that the potential for a model to be aware it was being tested created “significant limitations” for Anthropic’s ability to assess model safety. 

There are many examples of dangerous autonomous behaviour by AI. For example, OpenAI’s agents have hacked dozens of websites including AI startup Hugging Face and Australia’s health care system. OpenAI is delaying the release of its latest GPT iteration due to safety concerns. As AI models become smarter, concerns will only grow. Many domain experts assess AI as more dangerous than nuclear weapons due to its ability to become autonomous. The United States (US) administration is taking note of the dangers of what US President Donald Trump now calls “superintelligence”, and suggesting the development of safety standards. But what those standards could be and how long it will take AI to “jailbreak” them are big question marks. AI’s astonishing capabilities are also what lead to astonishing valuations. Anthropic plans capital expenditures of $518 billion on cloud and compute infrastructure. Sector analysts say AI-related investment will absorb over $10 trillion in capex in the next five years. That’s more than twice India’s gross domestic product. Only financiers with very deep pockets can provide this amount and, presumably, they will be seeking big returns on capital. That is an indication of how capable AI is expected to become. 

If Anthropic does hit $2 trillion, this would make it the seventh-most valuable listed company. It would overtake SpaceX (market cap of $1.95 trillion) and the SpaceX IPO is worth discussing in this context because it had both AI and Anthropic connections. In May, SpaceX, which has $37 billion in accumulated losses, listed at a valuation of over $1.8 trillion. That was in part because SpaceX is the holding company for xAI, the subsidiary under which Elon Musk holds group AI assets. Anthropic pays SpaceX $1.25 billion a month to rent compute resources. Anthropic’s financials (as reported) are fascinating. It had $4.6 billion in 2025 revenues. That was 12 times what it had made in 2024. It has already logged over $16 billion in revenue in the first two quarters of 2026. In 2025, Anthropic spent $12.6 billion on operations, with $8 billion in operating losses. Another $34 billion was written off. However, the growth is impressive. So are the demonstrable advances made by AI in the past three years. The fears of the damage it can do are realistic. Whether the valuations can be justified, or they are a gigantic bubble, is however, an open question.