Anthropic IPO: Claude, agentic AI and the opportunities and risks of the AI listing | CMC Markets
Anthropic IPO: Claude, agentic AI and the opportunities and risks of the AI listing
Andreas Lipkow
Chief Market Analyst
Anthropic could become one of the most important AI IPOs as investors weigh Claude's growth potential against rising compute costs, competition and valuation risk.
Anthropic’s potential stock market listing could become one of the most significant IPOs of the current AI wave. It is still unclear when Anthropic shares might actually list, how many shares could be offered or at what price the deal might be placed. Even so, the company is already attracting interest from retail investors. An IPO would offer direct access to one of the few large independent developers of so-called frontier AI models, while Anthropic shares are already trading in the pre-IPO market.
At the same time, the central question is whether the company’s exceptional growth momentum can justify what could be a very high valuation. Anthropic is growing at remarkable speed, but the development and operation of its AI models also require exceptional amounts of capital and computing power.
Founded in 2021, Anthropic has become one of the best-known developers of generative AI models alongside OpenAI and Google. Its most important product is Claude. The model family competes with OpenAI’s GPT and Google’s Gemini, among others, and is increasingly used by companies for software development, data analysis, document processing and other knowledge-intensive tasks.
Anthropic is particularly well positioned among enterprise customers and software developers. Products such as Claude Code are designed to integrate AI directly into programmers’ workflows. Strategically, this is about much more than another chatbot: Claude is intended to develop into core AI infrastructure for businesses.
Exceptional growth meets enormous costs
The company’s growth so far is one of the most striking developments in the AI sector. Anthropic said in May 2026 that annualised run-rate revenue had surpassed $47bn. At the end of 2025, the company said that figure had been around $9bn. In April, Anthropic also reported that more than 1,000 business customers were each spending more than $1m on an annualised basis on its products.
For investors, however, the distinction between run-rate revenue and actual realised annual revenue is crucial. Run-rate revenue annualises the current revenue level and is not the same as the revenue generated during a financial year.
According to the IPO prospectus, revenue increased twelvefold in 2025 to almost $4.6bn. At the same time, the company generated an operating loss of more than $8bn. Spending on computing power and infrastructure alone tripled to $7.33bn. The net loss of almost $42bn also included an accounting charge of around $34bn related to the revaluation of financing arrangements.
This means that, alongside growth, the capital intensity of the business model is increasingly moving into focus.
The valuation has also reached extraordinary levels. In May 2026, Anthropic raised $65bn in a Series H funding round and was valued at $965bn afterwards. Only a few months earlier, the valuation had been $380bn. A potential target valuation of up to $2tn has been discussed for the planned stock market listing.
For investors, one important principle follows from this: a technologically outstanding company is not automatically an attractively valued stock.
From AI model to enterprise infrastructure
Anthropic’s development needs to be viewed in the context of the wider AI investment cycle. The first phase of the boom was shaped above all by hardware. Companies such as Nvidia benefited from the explosive rise in demand for GPUs. Attention then shifted increasingly towards data centres, power supply, networking technology and cooling.
The next phase is likely to be shaped increasingly by the question of which companies can profitably monetise this vast infrastructure. This is where providers such as Anthropic, OpenAI and Google come in. They are trying to turn ever more powerful models into recurring revenue through subscriptions, APIs, enterprise solutions and autonomous AI agents.
The enterprise business is an important part of Anthropic’s strategy. The company is investing $100m in its Claude Partner Network, through which consultancies and IT service providers are expected to help implement Claude at businesses. At the same time, Claude is available through Amazon Web Services, Google Cloud and Microsoft Azure.
This multi-cloud strategy could prove an advantage. Enterprise customers often want to avoid becoming entirely dependent on a single cloud or AI provider. Anthropic, by contrast, can offer its models across different infrastructures and use different chip architectures.
Agentic AI could become the key growth driver
Perhaps the most important long-term growth driver is agentic AI. Until now, AI models have mostly been used to answer questions, create text, analyse data or generate software code. The next stage of development is for AI systems to carry out tasks increasingly independently.
Such agents could develop software, handle customer queries, review documents, analyse data or support complex business processes. If companies begin deploying thousands of these digital agents, Anthropic’s business model could change fundamentally.
Instead of occasional chatbot use, there would be continuous consumption of computing power and AI services. This could create a highly scalable enterprise business for Anthropic.
Computing power becomes the decisive cost factor
The other side of this growth story is the enormous infrastructure cost. Anthropic has committed to spending more than $100bn on AWS technologies over ten years. The agreement is intended to provide access to as much as five gigawatts of additional computing capacity. Further multibillion-dollar compute agreements have also been signed.
This highlights one of the central problems facing the entire AI sector: revenue growth alone is not enough. Anthropic must show over the long term that the economic value of its AI services is rising faster than the costs of GPUs, data centres, electricity and cloud infrastructure.
Unlike traditional software, almost every generative AI query creates additional compute costs. The more complex the models become and the longer autonomous agents work, the higher these inference costs can become.
For investors, the decisive issue will therefore not only be how quickly Anthropic grows. At least as important will be how gross margin, operating margin and free cash flow develop with each additional dollar of revenue.
Intense competition for models, capital and infrastructure
Competition is also exceptionally intense. Anthropic is competing with some of the world’s best-capitalised technology companies. Google is developing Gemini and has its own cloud infrastructure and TPU chips. Microsoft is closely tied to OpenAI and is integrating AI deeply into Azure and its software products. Amazon, meanwhile, is both a strategic partner and investor in Anthropic while also developing its own AI offerings.
Open-source models are also becoming more important. If powerful AI models become increasingly interchangeable, significant pricing pressure could follow. Anthropic therefore needs to build durable competitive advantages through model quality, safety, enterprise integration and a growing ecosystem.
Dependence on cloud and chip providers also remains a risk. Anthropic is trying to reduce this by using different platforms, including AWS Trainium, Google TPUs and Nvidia GPUs.
The AI market is therefore increasingly becoming a business in which technological quality alone is not enough. Access to capital, energy, data centres and semiconductors is becoming a strategic competitive advantage in its own right.
Regulation remains an important risk factor
Anthropic has focused heavily on AI safety and interpretability since its founding. This could help build trust, particularly among companies and government institutions.
At the same time, regulatory scrutiny is increasing as models become more powerful. Data protection, copyright, cybersecurity, liability and the possible misuse of powerful AI systems could create additional costs or restrict certain applications. If Anthropic becomes a listed company, these issues are likely to receive even greater attention from investors and regulators.
Opportunities and risks for investors
For retail investors, this creates an unusually wide opportunity-risk profile. Anthropic belongs to a small group of companies at the technological frontier of generative AI. Claude is gaining traction among enterprise customers and software developers, while strong partnerships with some of the world’s largest technology companies support further expansion.
If AI agents become a permanent part of business processes, the addressable market could expand substantially. An IPO would also give investors direct public-market access to a large independent frontier AI company for the first time.
Set against this are significant risks. The high private valuation already assumes extraordinary growth expectations. At the same time, developing and operating the models requires enormous investment. Competition with OpenAI, Google and other providers, possible regulatory restrictions and the still-open question of what long-term margins frontier AI companies can achieve add further uncertainty.
Valuation will be decisive at IPO
The valuation at Anthropic’s IPO will therefore be especially important. Even an excellent company can become a difficult investment if the issue price is too high.
Key information is still missing for a fully informed assessment. Investors should pay particular attention in the full IPO documents to actual realised revenue, revenue growth, gross margin, operating losses, free cash flow, long-term compute commitments, share-based compensation and the use of IPO proceeds.
The Anthropic IPO could therefore become an important test for the next phase of the AI boom. While the first big stock market winners primarily supplied chips and infrastructure for artificial intelligence, the market is now turning towards the economic monetisation of the models themselves.
Claude has become established particularly among enterprise customers and software developers. If AI models and autonomous agents eventually become infrastructure as important as cloud computing, Anthropic could become one of the central platform companies of that development.
For investors, however, it is not enough to believe that Anthropic could be one of the winners of the AI boom. The decisive question is what price the capital market demands for that future growth. The ambiguity is already visible in pre-IPO trading in the shares, where a degree of enthusiasm appears to be directly reflecting developments in the AI sector.
Only once reliable information on revenue, margins, cash flow, investment commitments and IPO valuation is available will it be possible to assess the opportunity-risk profile of Anthropic shares more thoroughly. Until then, the planned listing remains one of the most interesting potential capital market events in the AI sector.
