Anthropic is preparing for what could be one of the largest technology IPOs on record, but its race with Alibaba’s cloud unit is raising a question that bankers and investors are struggling to answer: what exactly is the moat, and is it worth nearly a trillion dollars? Fortune examined the dilemma this week, and the analysis has circulated widely among investors weighing whether to participate in the offering.
The stakes are enormous. Anthropic has been in discussions with bankers about a listing that could value the company at close to $1 trillion, according to people familiar with the matter, a figure that would place it among the largest companies in the world before it has a full year of profitability to show for it. The valuation is built on the assumption that frontier AI models are durable assets: hard to build, hard to replicate, and essential to the future of enterprise software.
The Alibaba competition is what tests that assumption. Alibaba’s cloud unit has been shipping open-weight models that close most of the performance gap with Anthropic’s Claude line, and it prices them aggressively, in some cases below the cost of running a comparable closed model. If a competitor can reproduce the capability with a fraction of the research budget, investors ask, then what part of the trillion-dollar value is actually protected? The question is not hypothetical: Alibaba’s models are already being deployed by enterprises that might otherwise have bought Anthropic’s.
The comparison is uncomfortable for another reason. Anthropic’s advantage has always been described in terms of safety and trust, the argument that enterprises will pay a premium for models built by a company that treats risk seriously. But safety claims are increasingly being matched, as the recent benchmark results from Chinese and Japanese labs show, and enterprises that test the models side by side are finding the practical differences smaller than the marketing suggests.
Anthropic’s answer, according to people familiar with its strategy, is that the moat is not any single model but the system around it. The company has built tools for evaluation, monitoring and deployment that enterprises need to run AI in production, and it has invested in the partnerships, including a deep relationship with Amazon, that give its models distribution. The argument is that customers are not buying a model; they are buying a platform, and the platform is harder to copy than the weights.
Investors are split. Some argue that the enterprise market is large enough to support premium pricing for a trusted provider, and that Anthropic’s revenue growth, which has compounded quickly, justifies a valuation that looks extreme only against traditional software metrics. Others point to the history of technology: every category leader in AI has eventually faced a lower-cost competitor, and the winner in those contests has usually been the one that owned distribution, not the one with the best research.
The IPO itself will force a resolution. Underwriters will need to set a price, which means committing to a view on defensibility, and the roadshow will be dominated by questions about Alibaba. Bankers involved in the process say the comparison has already reshaped the offering: materials emphasize the enterprise relationships and safety infrastructure over raw model capability, a shift that reflects where the company believes its durable value lies.
There is also a geopolitical layer. Alibaba’s challenge is partly a function of China’s industrial policy, which has poured resources into open-weight models as a way to spread AI capability widely and cheaply. American investors who would never buy a Chinese model are still affected by its price, because every enterprise that adopts an open model is one less customer for the closed providers. The competition is not just between two companies; it is between two models of how AI gets distributed.
For now, Anthropic is pushing forward. The company has continued to hire, to sign enterprise deals and to expand its product line, and its revenue is growing from a base that makes the trillion-dollar figure less fanciful than it would have been a year ago. But the Alibaba comparison has done something important: it has moved the IPO debate from whether the company can grow to whether the growth can be defended. That is a harder question, and it will not be settled by the roadshow.
The valuation debate has also been shaped by the market’s treatment of peers. OpenAI’s own valuation has climbed through successive funding rounds, and investors have shown an appetite for AI names that would have seemed reckless two years ago, a pattern that works in Anthropic’s favor. But public markets have been less forgiving than private ones: several AI-linked stocks have traded down sharply after earnings, and the reception of SpaceX’s volatile debut has reminded fund managers how quickly sentiment can turn. An IPO priced near a trillion dollars would arrive into that mood, and the underwriters know it.
