Anthropic's $100 Billion IPO: Is AI Gold or a Bubble?

Anthropic could hit the public markets as soon as next month. The AI company behind the Claude chatbot has a private-market valuation of nearly $1 trillion and has raised over $130 billion in cumulative funding. This IPO is expected to raise $100 billion, potentially putting it among the highest-valued listed companies in the world.

This IPO is far more than one company's debut. Anthropic and its chief rival OpenAI sit atop a vast ecosystem: computing infrastructure providers, chipmakers, energy companies, and AI startups have all spent the past four years betting that demand for AI products will keep climbing.

The market needs answers: Can demand for AI services justify the massive spending on chips, data centers, and infrastructure? Or will Anthropic's results prove that AI growth has a ceiling?

“For the entire market, it's a bellwether,” said Umesh Padval, an investor at Seligman Ventures with nearly 40 years of experience in the tech industry.

The market is on edge

Tech stocks have broadly weakened this month, with AI chipmakers leading the decline. The Philadelphia Semiconductor Index has fallen 26% from its June high, entering bear market territory. Over the past two weeks, Microsoft canceled its partnership with data center provider Nscale. A Gartner report predicts that 80% of the new cloud providers supplying computing power and data resources to AI companies could go out of business. Ken Griffin, billionaire founder of hedge fund Citadel, said bluntly that investors in AI are “almost certainly in a bubble.”

Behind the string of negative signals is the staggering scale of AI infrastructure spending. Amazon, Meta, Google, and Microsoft plan to invest at least $740 billion combined this year in AI infrastructure such as data centers. PitchBook, which tracks investment data, counts that private-market investors have poured $1.1 trillion into AI companies including Anthropic and OpenAI since 2022.

The public market's dependence on AI is also deepening. Eight of the top ten components of the S&P 500 are tech companies, and the top ten weighted stocks now account for more than 40% of the index, up from less than 20% a decade ago. If Anthropic is added to the index, it could displace Berkshire Hathaway or Eli Lilly—the only two non-tech companies with trillion-dollar valuations—to break into the top ten.

“Everyone is betting on this IPO. If it goes wrong, the whole market could be dragged down,” said Brad Gastwirth, global head of research and market intelligence at supply chain services firm Circular Technology.

Signs of slowing growth

This month, Anthropic showed investors financial figures at its San Francisco headquarters: third-quarter revenue of $16.8 billion, up 45% quarter over quarter. But growth clearly slowed from the second quarter's 141% pace.

How will the market read this slowdown? If Anthropic is losing share to OpenAI and low-cost open-source models, that would suggest overall demand for AI services remains solid. If it's a sign that industry-wide demand is contracting, the market reaction will be far more severe. Anthropic declined to comment.

Over the past year, Anthropic has been the hottest company in Silicon Valley. Claude Code, an AI coding tool for programmers, transformed the way coding gets done, sent revenue soaring, and pushed its private-market valuation to $965 billion—above OpenAI.

To handle the demand for its AI services, Anthropic has signed long-term computing lease agreements worth hundreds of billions of dollars, buying computing power from Google, Amazon, and Microsoft, chips from Nvidia, and data center capacity from Nscale. Its service providers, in turn, keep pouring money into their upstream suppliers, compounding the cycle.

But this summer, some customers began to feel token costs were too high and started looking for cheaper alternatives. Tokens are the billing unit for AI data processing and generation—the basis on which AI companies charge. Meta, one of Anthropic's largest customers, had expected to spend up to $10 billion a year but later cut back and switched to other AI models.

The AI IPO wave has already cooled off

AI companies that went public this year have seen lackluster stock performance. Cerebras, a Silicon Valley AI chipmaker, is down about 40% from its May IPO price. Accelevation, an AI infrastructure company, has fallen about a third from its September IPO price of $18. Musk's SpaceX, which listed in June, has dropped sharply from its highs, though it remains about 9% above its IPO price.

The IPO market overall is cooling. Holtec, which provides nuclear power support for the AI industry, withdrew its listing application last month. SoftBank's data center unit SB Energy and Aggreko, which powers data centers, have both delayed their IPO plans.

“AI infrastructure is the core theme of this year's IPO market, and it's also the segment hardest hit by the recent sell-off,” said Matthew Kennedy, senior strategist at Renaissance Capital, which tracks IPO activity.

Investment banks are already warning companies that valuations like those expected two months ago are hard to come by now. Goldman Sachs CEO David Solomon said on this month's earnings call that the IPO window hasn't closed, but companies hoping to get the valuations expected this summer will most likely have to wait. “The market weakened noticeably in September.”

Valuations are already being marked down

The market had originally expected Anthropic's IPO valuation to reach as high as $2 trillion. But amid slowing growth and a weak market, people familiar with the matter say some investors are already discussing a range of $1.5 trillion to $1.7 trillion.

AI industry executives are trying to calm nerves. TSMC CEO C.C. Wei said this month that the chip giant's profit growth has no ceiling.

Over the past weekend, concerns about AI spending continued to build, and the broader market fell again, with the Nasdaq posting its biggest single-day drop since June.

Now, all eyes are on this stock. The price it commands will determine whether every player in the AI supply chain can keep telling this story next year.

Source: cnBeta (via Global Market Broadcast)