• USD 2tn valuation; USD 200bn revenue projections need AI hype to grow
• CFO ROI concerns, open-source competitors could smash revenue targets
• Coming price war with fellow hyperscalers may further limit projections

Anthropic is gearing up for what is undoubtedly the most anticipated IPO for the rest of 2026, likely exceeding the enthusiasm generated by SpaceX in 1H.

It is the arguably the creator of the most powerful and sought after AI models on the planet, led by philosopher CEO Dario Amodei, considered a modern-day Prometheus in many circles, seeking to bring artificial enlightenment to mankind.

Talk of Anthropic’s huge IPO ambitions, however, brings to mind the tragedies of ancient Greek myth, when hubris, no matter how well-intentioned, angered the capricious and unforgiving gods.

Like its interstellar, Elon Musk-led AI peer, Anthropic is reported to be seeking a gargantuan valuation at IPO, potentially topping USD 2tn, and exceeding SpaceX’s listing market cap.

Anthropic was reported to have an annualised revenue run rate of USD 65bn for 2026 as of the end of July, representing a 30x price/sales multiple for 2026 at the aforementioned valuation.

Given Anthropic’s revenue of USD 10bn across all of 2025, a 200x P/S multiple for its 2025 numbers, the AI stalwart is already showing the success of its models and the efficiency of its delivery to enterprise clients.

The company also reportedly projects revenue of USD 200bn for 2028, meaning that an IPO investor will be asked to value the business at just 10x its 2028 numbers.

Assuming revenue of around USD 65bn in 2026, for Anthropic to hit its 2028 target it will need to grow its revenue 3x by the end of 2028.

Given it is on track to grow its sales by 6.5x from the end of 2025 to the end of 2026, many investors will be happy to take the bet that its incredible momentum will continue.

It seems a sure-fire thing, if we are not hitting the peak of not just the AI cycle, but Anthropic hype as well.

Anthropic is clearly a spectacular business. Its agentic AI models have begun to transform processes across white collar industries, producing much-touted individual efficiency gains. But, at a top-line level, there is still some debate over whether these efficiencies have produced a tangible return on investment given the huge cost of AI tokens.

In an agentic world where Claude-powered agents automate the entire white-collar services and enterprise software industries, USD 2tn might almost seem cheap.

But there has been some pushback which might indicate we are hitting the peak of the frontier LLM cycle.

While a Claude-powered financial agent can knock up a pitch deck at a speed that is unimaginable to mere mortal bank analysts, it is arguably still far cheaper to get a human to do it, especially given the need to check anything client facing with a fine-tooth comb for a myriad of potential hallucinations.

This is not to say that Anthropic’s products are not amazing or arguably the best and most powerful tools on the market – they are – but they are also perhaps the most expensive.

And this focus on AI-cost and performative “tokenmaxxing” has come at a time when models are becoming increasingly commoditised.

Far cheaper open-source models from China, such as DeepSeek, Qwen and Kimi, alongside US tech giant Meta’s Muse and Llama Series will continue to provide clients will more budget friendly options to Claude. Even among fellow hyperscalers, June saw Alphabet fire an opening shot in what is likely to become a price war, cutting its monthly AI Plus plan to USD 4.99 per month from USD 7.99 and doubling storage. Last week it rolled out Gemini 3.7 Flash – a go-to coding and agentic AI “workhorse” – while temporarily halving API costs.

The better they get, the more attractive these solutions become as an alternative for ROI-obsessed CFOs.

Anthropic’s continued mastery of its sector depends on hype for AI, and namely for its Claude models, continuing to grow apace. However, potential investors would do well to give a quick glance to the increasingly cited Gartner Hype Cycle that is designed to discern the difference between hype and commercial viability.

The first phase of excitement ends at what Gartner calls the “Peak of Inflated Expectations” when early adopters make visionary and sometimes outlandish claims about a technology and what it can do.

We still seem to be close to the peak now for both Anthropic and the potential of artificial intelligence.

But in Gartner’s model, the peak is followed by a “Trough of Disillusionment” when “early tests and real-world uses fail to meet the high, unrealistic promises made during the peak hype phase.”

This is the real fear around frontier LLMs: at some point enterprise clients will abandon these high-cost investments when the promises of vast efficiencies, mass automation and higher profits fail to materialise.

In this most bearish of scenarios, LLMs are fully commoditised as DeepSeek, Qwen, Meta’s Llama and others drive intelligence costs to almost zero.

This would take place at the same time core enterprise clients determine that while AI is hugely useful for several tasks, autonomous agents are still too unreliable for mission-critical client facing tasks, running the risk that CFOs will slash AI software budgets and pivot to cheaper LLMs.

In this scenario Anthropic could hit a wall soon in the years after IPO, causing revenue growth to slow.

This would force it to suit up in the coming pricing war with its biggest rivals, Google parent Alphabet and OpenAI, all now competing for a smaller slice of premium enterprise AI clients still seeking highly sophisticated models more powerful than the cheaper LLMs used across most of the corporate client base.

That’s likely not a large enough total addressable market (TAM) to make Anthropic’s USD 200bn 2028 revenue projections feasible.

Anthropic also would be unable to increase its free cash flow margins given the huge spend in compute required to train the most sophisticated models. In that world, a USD 2tn valuation might seem very rich indeed, making it a prime target for aggressive short sellers.

There is no guarantee here that this happens and it is certainly not even a base case.

As said earlier, Anthropic is at the pinnacle of AI model developers capable of aiding exceptional innovation that could, like Promethean fire, be a genuine benefit to humanity, not just white-collar services. There should always be value in investing in that sort of innovation.

But by setting such a huge valuation target at IPO, it might just be creating its nemesis which would bring the fury of the equity market gods down upon it, despite any noble ambitions.