Anthropic is preparing for what could be one of the largest IPOs in the history of the technology sector. The company has already filed for an initial public offering and, according to sources, may go public as early as October. At the same time, Anthropic’s potential valuation could exceed $2 trillion, potentially making it one of the largest IPOs on record.
The problem is that investors are being asked to value a company that does not yet have a financial base to support such a market capitalization. The valuation is based not on current performance, but on forecasts for several years ahead. Anthropic expects its revenue to reach $190-200 billion in 2028. For comparison, the company’s estimated annual revenue in May was about $47 billion. Thus, in less than three years, Anthropic expects to roughly quadruple the size of its business.
For a rapidly growing technology company, such a valuation methodology is not in itself extraordinary. However, in the case of Anthropic, investors are effectively being asked to buy into not the current business, but a significant portion of its future growth.
The company forecasts second-quarter 2026 revenue of no less than $10.9 billion, more than double the level of the previous quarter. Moreover, Anthropic expects to achieve a quarterly operating profit of $559 million for the first time. Its estimated annual revenue has grown more than tenfold each year over the past three years.
This pace of growth allows investors to overlook the scale of its spending. Anthropic needs to constantly increase its computing capacity and spend heavily on model training, inference, and workforce expansion. The company’s entire investment story is based on the assumption that revenue will continue to outpace expenses and that technological improvements will reduce the cost of running AI models.
Anthropic is simultaneously demonstrating increasingly impressive model capabilities while facing an increasingly apparent challenge: how to control them. The company recently announced the existence of an experimental Model 2, which surpasses its previous flagship, Mythos. Anthropic does not currently plan to make it widely available, as the model’s capabilities already show signs of rapid advancement in autonomous research, software development, and agentic tasks.
At the same time, the company was forced to raise its own assessment of the potential risks of uncontrolled model behavior from very low to simply low. The wording doesn’t sound overly dramatic, but for a business aiming for a trillion‑dollar market cap, it matters. The more powerful the models become, the more difficult it is to maximize their commercial value while controlling the potential consequences of their use.
The company has also decided to implement special statistical watermarks in Claude’s responses, designed to indicate the likelihood of AI involvement in text creation. The technology does not require additional tokens and is not expected to significantly affect the quality of responses; however, some users have already started canceling their paid subscriptions.
The issue is particularly sensitive for developers and companies. Users fear that watermarks could raise questions when delivering software code to clients, preparing documents, or using AI in academic work. Some former subscribers have already said they are switching to competitors, including Grok, a shift that could benefit SpaceX’s AI business and potentially provide a positive signal for SpaceX stock.
For Anthropic, this does not pose an immediate threat to its business, especially given the scale of its current growth. But ahead of Anthropic IPO, such signals become much more important. Investors will have to determine how loyal the customer base will remain if the company begins to unilaterally change the rules governing how its products can be used.
As a result, Anthropic is approaching its stock market debut in a somewhat paradoxical position. On the one hand, the company is demonstrating explosive revenue growth, rapidly approaching operating profitability, and developing technologies that the market considers among the best in the industry.
On the other hand, its estimated valuation of $2 trillion requires virtually flawless execution of a multiyear plan. Anthropic must simultaneously increase revenue roughly fourfold over several years, make computing costs more manageable, retain customers, withstand competition, and learn to control models whose capabilities are growing faster than the rules governing their use can evolve.
Therefore, the main question surrounding Anthropic is not the IPO itself. What matters more is how much of the growth investors are being asked to price in today will ultimately translate into actual revenue. If the forecast of $190-200 billion in revenue for 2028 comes true, today’s valuation may turn out to be quite reasonable. However, if the growth rate slows even modestly, investors may have to reassess a business they are being asked to value in the trillions of dollars.
