The company behind Claude is preparing for one of the largest listings in history. Revenue is surging, but investors need to answer three questions: how sustainable is the growth, how much will it cost, and how much should they be willing to pay for it?
Artificial intelligence has already changed the way businesses operate. Now it must prove that it can deliver returns commensurate with the expectations built into its valuations.
Anthropic, the company that develops Claude, is at the centre of this test. With the filing of an S-1 registration statement for its Nasdaq listing on 26 October 2026, the market has a rare opportunity to examine the financials of one of the most important independent developers of advanced AI models.
Interest extends beyond the scale of the impending public offering. Reports have put the valuation at $1.5tn–$2tn, with the capital raise potentially reaching $100bn. These figures, however, are expectations, not final offering terms.
The real question is a different one: can a company turn AI’s explosive growth into sustainable profitability when it needs to commit hundreds of billions of dollars to computing power and infrastructure?
The answer matters for far more than one stock. It has implications for technology company valuations, semiconductor suppliers, investment in data centres and energy, and the ways businesses and economies will use AI.
1. What exactly is the investor buying?
Anthropic was founded in 2021 by Dario Amodei, his sister Daniela Amodei and other former OpenAI researchers. Its core product is the Claude family of models, which serves individuals, businesses and developers.
Its business is not limited to a digital assistant. It includes coding tools, the automation of complex tasks and the integration of AI into business applications.
| Product | Business function |
|---|
| Claude | Analysis, research, writing and decision support |
| Claude Code | Software development, testing and refactoring |
| Claude Cowork | Carrying out multi-step knowledge work |
| Claude API | Integrating AI into applications and business systems |
| Claude Enterprise | Business use with controlled access and tool integration |
The economic significance of this shift is that AI is moving beyond experimentation and finding applications in day-to-day business. Customers are no longer paying simply to try a new technology. They are building it into their processes, aiming for lower costs, faster task completion and higher productivity.
This creates a real market. It does not, however, guarantee that every company operating in it will deliver comparable investment returns.
2. Impressive growth; its staying power remains to be proved
Anthropic’s financials show an exceptionally rapid acceleration in revenue.
| Period | Revenue |
|---|
| 2023 | $41m |
| 2024 | $386m |
| 2025 | $4.6bn |
| First half of 2025 | $1.16bn |
| First quarter of 2026 | $4.73bn |
| Second quarter of 2026 | $11.6bn |
| First half of 2026 | $16.3bn |
Source: Anthropic, Form S-1, SEC.
Revenue in the first half of 2026 was about 14 times higher than in the same period of 2025. The second quarter alone generated more revenue than the whole of 2025.
This suggests that demand for advanced AI models is more than a stock market narrative. Customers are paying, and businesses are increasing their use of the technology further.
There is, however, an important distinction: growth from a small starting base cannot be repeated indefinitely. As revenue rises, it becomes harder to sustain the same growth rates.
Moreover, revenue from AI services depends significantly on usage. The company may attract more customers, but growth will prove sustainable only if they continue to use its services, pay the prices required, and find enough economic value to justify the expense.
The investment takeaway: rapid revenue growth shows that real demand exists. On its own, however, it does not prove that the current pace of growth can be sustained.
3. The $80.7bn loss and the substance of profitability
One of the most striking figures in the registration statement is the net loss of $80.7bn for the first half of 2026. Interpreting it, however, requires care.
| Financial measure | First half of 2026 |
|---|
| Revenue | $16.3bn |
| GAAP net loss | $80.7bn |
| ↳ From the revaluation of convertible securities and related liabilities | $72.4bn |
| Operating result | -$693m |
| Adjusted operating profit | +$1.62bn |
| Second-quarter operating profit | +$974m |
Source: Anthropic, Form S-1, SEC. Adjusted figures are not equivalent to GAAP results.
Most of the net loss is linked to non-cash accounting changes in the fair value of convertible securities and other financing instruments. An increase in the company’s valuation can further increase the accounting charge on these liabilities without causing a corresponding immediate cash outflow.
This does not mean the loss is irrelevant. It means that accounting losses should not be confused with cash burn.
The operating picture is clearly better, though not without caveats. The operating loss in the first half was much smaller than in the corresponding period of 2025, while the second quarter recorded an operating profit.
Three measures are particularly important for investors:
- Operating profitability: can the company cover its operating costs without relying constantly on new funding?
- Cash flow: is revenue converted into cash after necessary investment?
- Infrastructure costs: is the profit margin high enough to fund the next phase of growth?
The distinction is essential. A business may report an operating profit but still need enormous amounts of capital to expand its infrastructure. This is also the central question for the so-called AI trade as a whole.
4. The real test: how much does AI cost?
Artificial intelligence requires computing power, specialist semiconductors, data centres, electricity and network infrastructure. Increased usage can drive up revenue, but it also raises costs.
The registration statement sets out infrastructure commitments on a massive scale.
| Item | Reported amount |
|---|
| Existing commitments to computing providers | About $417bn |
| Additional orders for 2028–2031, with pricing not yet finalised | $125–150bn |
| Commitments under eight infrastructure agreements | About $657bn |
| Non-cancellable portion of the latest commitments | About $498bn |
Source: Anthropic, Form S-1, SEC. The figures cover different categories of contractual commitments and should not be mechanically added together, as there may be overlap.
The agreements include long-term access to infrastructure from large technology companies. This scale illustrates how capital-intensive the development of advanced AI models is becoming.
The problem is not that infrastructure investment is necessarily excessive. It is that it commits the company to future spending that must be covered by revenue that is not guaranteed.
If usage continues to grow, the infrastructure can support higher revenue and a competitive advantage. If growth slows, the same commitments could constrain the company’s profitability and flexibility.
This is the fundamental asymmetry of the business model: demand can change quickly, but a large part of the costs has been contracted for years.
The key question is not only how much revenue AI generates, but how much the company retains after the cost of computing power and committed infrastructure.
5. Dependence on partners
Anthropic works with major cloud platforms, including Amazon Web Services, Google Cloud and Microsoft Azure. According to the registration statement, 48% of revenue in the first half of 2026 came from these three platforms.
The partnerships provide access to infrastructure, customers and international distribution. At the same time, they create concentrated risk.
The major cloud providers are not just suppliers. They are also investors in Anthropic, as well as its competitors, as they develop their own AI models and services.
The relationship runs both ways. Anthropic needs their infrastructure and distribution channels, while the providers have an incentive to increase usage of their services. Their priorities, however, will not always align.
For investors, there are three main risks:
- Bargaining power: the terms of access to computing power and distribution can affect profit margins.
- Competition: the same partners have the resources to develop competing products.
- Revenue concentration: dependence on a limited number of major channels and customers such as SpaceX could make the company more vulnerable to changes in commercial policy.
Anthropic does not need to stop working with the major providers. It does, however, need to retain enough technological differentiation and bargaining power to avoid becoming overly dependent on them.
6. Risks that do not show up in a valuation multiple
The valuation of an AI company depends on more than growth, profit margins and capital expenditure. The registration statement itself highlights broader risks that could affect the business in the future.
Technology risk. A model’s current lead does not guarantee that it will last. Competition from OpenAI, Google and other providers could put pressure on prices or reduce product differentiation.
Pricing risk. If customers have more alternatives, Anthropic may need to cut prices even if the overall market continues to grow.
Regulatory and legal risk. Data use, intellectual property rights, model safety and restrictions on certain applications could lead to additional costs or limits on operations.
Safety risk. A serious incident involving the use or behaviour of an advanced model could undermine customer trust, increase compliance costs and lead to legal consequences.
Capital requirements risk. If funding needs exceed available cash flow, the company may need to raise more capital, potentially diluting existing shareholders.
These risks do not negate the investment opportunity. They do, however, show why growth in the market should not be confused with certainty about the returns of any particular company.
7. How expensive is a $1.5tn–$2tn valuation?
Without a final offer price and definitive listing terms, the valuation cannot be determined with the precision required for an investment decision. It is possible, however, to make an illustrative calculation based on revenue in the second quarter of 2026.
Revenue of $11.6bn equates to an annualised run rate of $46.4bn. This calculation assumes that second-quarter revenue is repeated for four quarters. It is not a forecast for the full year.
| Illustrative valuation | Multiple of annualised revenue |
|---|
| $1tn | 21.6 times |
| $1.5tn | 32.3 times |
| $1.75tn | 37.7 times |
| $2tn | 43.1 times |
Calculations based on annualised revenue of $46.4bn. These are not official offering terms. The valuation-to-revenue multiple is not a measure of profitability.
The figures show the scale of the expectations that a valuation of this size would embody. For a $1.5tn valuation to represent 15 times annual revenue, Anthropic would need to reach $100bn in annual revenue. That would be more than double the second quarter’s annualised run rate.
That does not mean the company cannot get there. It means investors are being asked to pay today for substantial future growth.
There is also a second dimension. Not all revenue is of equal value. If growth requires disproportionate investment in computing power, the same level of revenue could generate far less economic value than the multiple suggests.
That is why the valuation needs to be considered alongside the prospects for profitability, cash flow and future capital requirements. An outstanding business can prove a poor investment if bought at too high a price.
8. Governance: who will control the company?
A stock market listing does not mean that new shareholders will have corresponding influence over Anthropic’s strategy.
The proposed governance model provides for special classes of shares and a control structure under which the seven co-founders, through Founder LLC, retain 50.1% of the voting rights in general shareholder votes. In parallel, the Long-Term Benefit Trust has a special role in electing and removing a majority of the board, under the terms of the registration statement.
Anthropic is organised as a public benefit corporation, with an obligation to take into account both shareholders’ financial interests and its stated public benefit purpose.
This approach may help the company maintain its strategy over the long term. It does, however, limit ordinary shareholders’ ability to influence decisions they consider unfavourable, simply because they do not have the votes.
Investors therefore need to distinguish between two things: an economic stake in the company’s growth and the ability to influence its management. The stock may offer the first without providing the second to the same extent.
9. What it means for Greek investors
Anthropic matters to Greek investors even if they never buy its shares. The public valuation of a leading AI lab could influence expectations across the entire value chain: semiconductors, equipment, cloud, data centres, networks and energy.
| Investment exposure | What investors need to monitor |
|---|
| Global equity indices and technology funds | Technology weightings, valuations and concentration risk |
| Semiconductors and equipment | Whether demand for computing power translates into sustainable profits |
| Energy and infrastructure | Actual demand for electricity, investment and returns on capital |
| European and Greek software companies | Their ability to differentiate and retain customers in the face of large platforms |
| AI investments | The purchase price relative to potential future profitability |
Greece does not need to create the next global developer of foundation models to benefit from artificial intelligence. The more realistic opportunity lies in applying it in sectors where the country has activity, expertise or comparative advantages.
- Shipping: better maintenance forecasting, route optimisation and fuel management.
- Tourism: automated customer service, booking management and more effective pricing.
- Logistics and trade: demand forecasting, inventory management and reduced operating losses.
- Energy: forecasting supply and demand, better grid management and optimised facility operations.
- Financial services and public administration: process automation, document analysis and faster service, with appropriate security and data protection controls.
The prerequisite is that investment delivers a measurable economic result. Adopting AI does not in itself create a competitive advantage. That comes when the technology improves productivity, quality or costs relative to competitors.
For Greece, the challenge is to generate more added value from the use of technology, rather than limiting itself to buying services developed and priced elsewhere.
10. Before buying the stock: five questions
Anthropic’s listing could be a historic event for markets. That, however, is not a sufficient reason to buy.
Investors need to answer five questions:
- What is the final valuation? The difference between $1.5tn and $2tn is enormous in absolute terms and directly affects the growth rate required.
- What is the underlying profitability? Investors need to examine operating margins, cash flow and capital expenditure, not just revenue and adjusted profit.
- How much of the future cost is committed? Long-term infrastructure contracts can be an advantage in a growing market or a burden if growth slows.
- How strong is the competitive position? Technological differentiation, customer retention and the ability to maintain pricing will determine whether today’s growth turns into long-term economic value.
- What margin of safety does the price offer? Even with favourable prospects, the investment needs to withstand slower growth, pressure on margins and higher capital requirements.
The practical approach is clear: do not let excitement about the listing predetermine the decision. Examine the final terms, assess the price against realistic profitability scenarios and compare the expected return with alternative investments.
Conclusion: AI can change the world. The share price is another matter
Anthropic represents a new phase in the AI economy. Its growth shows that there is real business demand for advanced models and that companies are willing to pay for productivity and automation.
That growth, however, comes at a significant cost: high capital requirements, long-term infrastructure commitments, dependence on major providers and uncertainty over whether current rates of revenue growth can be sustained.
The issue takes on particular significance now that the recent results of the large technology groups (hyperscalers), which are investing huge sums in artificial intelligence, have revived questions about the returns on these investments and the resilience of their business models. It is no coincidence that Blackstone expects Anthropic’s bonds to be rated junk.
Against this backdrop, a potential Anthropic listing will be a crucial test for the market. Not only of the valuation it seeks, but above all of whether investors will be willing to fund that valuation, given such high expectations and growing doubts about the sustainability of investment in AI.
The forthcoming listing will test how far the market is willing to look ahead when valuing AI’s future profits. It will not, by itself, prove that any valuation is justified.
For investors, the distinction is fundamental: recognising a technological revolution is one thing; buying a company at a price that leaves room for error, delays and lower returns is another.
Anthropic may be one of the companies that shape the future of AI. The question for investors is not so much whether that future will arrive, but how much of it they have already paid for in advance.
Sources: Anthropic, Form S-1 — Registration Statement, 26 October 2026, U.S. Securities and Exchange Commission, EDGAR — Filing Details, Reuters, Should you invest in Anthropic? Five takeaways from its S-1, 28 October 2026, Fortune, Anthropic’s S-1 shows $16.3 billion in revenue and an $80.7 billion net loss, 26 October 2026.
Note: Valuations, multiples and annualised revenue calculations are illustrative. The reported terms of the public offering remain subject to change until formally announced.
Nicholas Havoutis has extensive experience leading strategic finance functions, having held executive roles at JPMorgan (New York), Chase Manhattan Bank (London) and Eurobank (Athens). He also has a significant presence in the media sector. Today, as head of SoZone Limited, he advises businesses and investors on international growth, organic optimisation, and mergers and acquisitions strategies.
