AI stocks: Is the easy money over? Aswath Damodaran warns investors of a new market reality ahead of Anthropic IPO

The artificial intelligence trade may be entering a more demanding phase, and investors who have so far benefited from buying almost anything associated with AI may need to rethink their strategy, according to experts.

Aswath Damodaran, professor of finance at New York University's Stern School of Business, believes the broad-based rally in AI stocks is beginning to lose steam, with investors increasingly differentiating between companies that stand to benefit from the AI boom and those that may struggle to turn the massive investments into profits.

In his November 9 post on Musings on Markets, titled The AI Trade Deconstructed: AI Builders, AI Suppliers and AI Product Firms!, Damodaran argues that the easy, broad-based gains from the AI boom may largely be behind investors. As the technology moves from hype and infrastructure spending towards actual business building, companies within the AI ecosystem are likely to experience increasingly divergent financial fortunes.


"I do believe that the easy AI trade, where you could buy any company that was part of the AI ecosystem (supplier, buyer or product) and make money from a general optimism about AI's potential is mostly done," he wrote.

For investors, the distinction matters. Companies supplying chips, building data centres, developing AI applications and ultimately paying for AI services may all be associated with the same technological revolution, but their revenue models, capital requirements and paths to profitability are fundamentally different.

The shift comes at a particularly sensitive time for markets, with the anticipated Anthropic IPO potentially becoming an important test of investor appetite for AI businesses.

ALSO READ: Anthropic eyes $1.4 trillion valuation in blockbuster IPO, targets up to $90.4 billion in fundraising

The AI trade is no longer one trade

Damodaran divides the AI ecosystem into four broad categories: AI suppliers, AI builders, AI product companies and AI customers.

AI suppliers include semiconductor manufacturers such as Nvidia, alongside companies providing electrical equipment, power infrastructure and real estate for data centres.

AI builders comprise hyperscalers such as Alphabet, Amazon, Meta and Microsoft, which are committing hundreds of billions of dollars towards developing the computing infrastructure needed to support AI.

The third category consists of AI product companies, including Anthropic and OpenAI, which use this infrastructure to develop and commercialise AI models, applications and services.

The fourth, and arguably most important, group is the customer base: individuals and businesses that ultimately pay for AI products and services.

According to Damodaran, the economics of these groups are beginning to diverge.

Infrastructure suppliers are among the first beneficiaries of the AI investment cycle because spending by hyperscalers immediately translates into revenues for them. Builders, meanwhile, must absorb substantial capital expenditure before their investments generate meaningful returns.

AI product companies face a different challenge. They need to convert growing adoption into recurring revenues and, eventually, sustainable profits.

Customers, in turn, must demonstrate that AI tools deliver enough productivity gains to justify the growing expenditure.

This creates a fundamental distinction between companies benefiting from AI spending today and those whose valuations depend on the eventual profitability of AI products.

Why Nvidia and hyperscalers could move in different directions

One of Damodaran's more important observations is that the financial interests of AI companies are not always aligned.

For instance, excessive investment in data centres could be positive for chipmakers and infrastructure suppliers, at least initially, because it increases demand for their products and services.

But the same spending could become a problem for hyperscalers if the infrastructure fails to generate adequate returns.

"Those common interests notwithstanding, these businesses also have competing agendas, and decisions and actions that affect one group negatively can create benefit for others," Damodaran noted.

This distinction could become increasingly important as investors begin scrutinising capital expenditure, cash flows and returns on invested capital rather than simply rewarding companies for their exposure to AI.

Damodaran also points to a potential shift in how markets value infrastructure suppliers.

As the initial data-centre construction boom matures, growth rates for chipmakers and other suppliers could moderate. New competitors, alternative supply sources and efforts by hyperscalers to develop their own capabilities could also put pressure on margins.

For builders, the investment equation will depend on how quickly completed infrastructure can be monetised.

For AI product companies, the challenge will be to establish pricing power and generate sufficient revenues to cover the cost of computing infrastructure, research and development, and employee compensation.

In other words, being part of the AI ecosystem may no longer be enough to justify a premium valuation.

The biggest risk: AI investment without profitable customers

Perhaps the most important part of Damodaran's argument concerns the demand side of the AI business.

Despite the extraordinary investment in AI infrastructure and the excitement surrounding frontier models, the ultimate economic case for the technology rests on whether customers are willing to spend substantially more on AI products and services.

Damodaran estimates that collective revenues from AI products and services are currently around $250 billion, a relatively small figure compared with the scale of investment made across the ecosystem.

The gap between infrastructure spending and monetisation is therefore a key risk investors need to monitor.

If AI products prove more useful than anticipated and customer demand expands rapidly, AI product companies could be the first major beneficiaries. Their growing revenues and pricing power could eventually translate into stronger profitability, benefiting infrastructure builders and suppliers as well.

However, if adoption falls short of expectations, the consequences could spread across the entire ecosystem.

AI product companies could struggle to meet their revenue targets while remaining committed to expensive computing contracts. That could affect hyperscalers leasing out infrastructure, which could subsequently reduce demand for chips, equipment and other data-centre inputs.

For investors, this makes customer adoption and monetisation arguably more important than headline AI spending figures.

Three scenarios that could determine the future of AI stocks

Damodaran outlines three broad scenarios for how the AI business could evolve, depending on whether customer demand meets, exceeds or falls short of expectations.

In the first scenario, AI products and services prove more useful than anticipated, with demand expanding faster than expected. AI product companies would benefit first through higher revenues and improved pricing power. As these businesses scale and become more profitable, the benefits would extend to infrastructure builders, which could expand capacity, and eventually to suppliers of chips, electrical equipment and other infrastructure.

In the second scenario, AI adoption and monetisation broadly meet expectations. In this case, AI product companies may hold their own in aggregate, but competition would determine which businesses emerge as winners. Stronger companies would gain market share, while weaker players could struggle or exit the market. Investors would increasingly differentiate between companies rather than rewarding the entire AI ecosystem.

The third scenario is more concerning. If AI products and services deliver less value than expected, customer demand and monetisation could fall substantially short of expectations. AI product companies would struggle to generate sufficient revenues to meet their commitments for computing capacity. Some could seek to renegotiate or break those commitments, affecting hyperscalers that have invested heavily in infrastructure. Builders could then reduce or abandon planned capacity, creating further pressure on chipmakers and other infrastructure suppliers.

The scenarios underline why the long-term fortunes of AI companies cannot be judged simply by how much money is currently being invested in the technology.

The Big Market Delusion: Why AI could still face a correction

Damodaran also draws attention to what he calls the "Big Market Delusion", a concept that he believes is particularly relevant to the AI investment cycle.

The idea is that expectations of an enormous addressable market, combined with overconfidence among founders, venture capitalists and corporate executives, can encourage excessive investment. Businesses build capacity in anticipation of future demand, while investors assign valuations that assume a large and highly profitable market will inevitably emerge.

Damodaran believes AI has already experienced elements of this phenomenon.

In his assessment, the industry has seen overinvestment by AI builders, excessive long-term computing commitments by AI product companies and overpricing by equity markets, at least in aggregate.

However, he does not believe that the possibility of a correction means investors should abandon AI altogether.

He argues that overinvestment and subsequent corrections are often features of periods of major technological disruption. Although the process can create significant financial pain, it can also be a necessary part of the development of transformative industries.

The correction, when it comes, is unlikely to affect all companies equally. Businesses with stronger economics, more sustainable revenue models and better competitive positions could emerge in a stronger position, while companies that have overestimated demand may struggle.

For investors, the distinction is between believing in the long-term potential of AI and assuming that every company associated with the technology will necessarily succeed.

The warning behind the AI correction

Damodaran also draws attention to the market's recent volatility.

He notes that AI stocks have already started displaying different performance patterns in 2026, even before the sharp sell-off in November.

According to his analysis of selected publicly traded AI companies, the group collectively lost nearly $1.8 trillion in market capitalisation on November 5 alone, with chip stocks bearing much of the damage.

Yet the market reaction was not uniform. While AI chip and electrical equipment companies suffered substantial losses, the hyperscalers collectively managed to increase their market capitalisation during the week.

Damodaran's analysis also suggests that AI companies had been experiencing divergent performance trends throughout 2026, with strength in some groups masking weakness in others.

He points out that AI companies, in aggregate, increased their share of overall US equity market capitalisation for much of the year, reaching approximately 37% on September 30, 2026. However, since the end of the third quarter, most AI groups have struggled, with AI real estate and construction being an exception.

The divergence could reflect investors becoming more discerning about the economics of individual AI businesses.

Damodaran suggests that concerns around data-centre construction could affect suppliers more immediately than builders. If political developments slow infrastructure expansion, companies supplying chips, electrical equipment and other inputs could be among the first to feel the impact.

Hyperscalers, whose shareholders have increasingly questioned the scale of AI capital expenditure, may even welcome a moderation in infrastructure spending.

The implication is clear: investors should not assume that a sell-off in one part of the AI ecosystem automatically makes every other AI stock a poor investment.

Anthropic IPO: A test of AI valuations?

The timing of Damodaran's observations is particularly relevant given the anticipated Anthropic IPO.

The company has become one of the most closely watched private AI businesses, and its public listing could offer investors a fresh opportunity to assess how much they are willing to pay for a fast-growing AI product company.

But Damodaran believes the recent deterioration in market sentiment could complicate the pricing exercise.

He refers to reports suggesting that Anthropic's October annualised revenue run rate reached $80 billion, although he cautions that such metrics need to be viewed in the context of the company's substantial infrastructure commitments and future profitability.

For a young AI company, impressive revenue growth does not automatically translate into an attractive valuation.

Damodaran's valuation sensitivity analysis illustrates the issue.

He examines the effect of increasing the assumed cost of future computing capacity by lowering the sales-to-capital ratio from 3 to 2.5. He also increases the assumed impact of regulation on profit margins from 5% to 7.5%.

Even while retaining a long-term revenue target of $1 trillion, these changes reduce his estimated equity value for Anthropic to $1.25 trillion.

The analysis highlights how sensitive the valuation of a young AI company can be to assumptions about infrastructure costs, regulatory exposure and future profitability.

Damodaran also points to political risks surrounding data-centre expansion, particularly following the November 3 election results. He notes that Texas, which accounts for approximately a third of new data centres expected to come online in the near term, could see slower infrastructure development. He also raises the possibility of tighter regulatory restrictions on AI, potentially creating additional pressure on margins.

While he does not expect the IPO to be withdrawn or delayed, he anticipates that pricing could move closer to the $1.25 trillion valuation level, potentially accompanied by a smaller share offering.

He also suggests that a smaller offering could give the stock greater scope for a price jump on its first trading day, helping generate positive market momentum.

The broader lesson for investors is that IPO valuations in the AI sector will increasingly depend on assumptions about future infrastructure costs, regulatory exposure and profitability, rather than revenue growth alone.

What should investors do now?

Importantly, Damodaran is not arguing that investors should abandon AI stocks altogether. Instead, he believes the next phase of the AI investment cycle will reward selectivity, valuation discipline and a willingness to distinguish between companies with different business models.

He cautions against both extremes: dismissing the entire AI phenomenon as a bubble and assuming that every company associated with the technology will inevitably succeed.

For existing investors, he suggests using market corrections to gradually reduce exposure rather than attempting to predict the exact timing of a downturn.

For those looking to enter the market, his approach is to identify businesses they genuinely want to own, determine the valuation at which they become attractive and establish buying triggers rather than chasing momentum.

He also makes an important point about the opportunities that emerge during periods of market uncertainty.

"Go where it is darkest" is the underlying investment philosophy he highlights, arguing that the payoff from fundamental analysis and valuation is often greatest when uncertainty is at its peak.

This does not mean buying every stock that has fallen sharply. Rather, it means identifying businesses whose underlying fundamentals remain attractive despite the market's negative sentiment and waiting for valuations to become compelling.

His conclusion is particularly relevant for investors navigating the next phase of the AI boom.

"Stay with your beliefs about AI but leave the door open for different opinions."

Disclosure: This article has been written by Kumar Gaurav, who is not a Sebi-registered Research Analyst or an Investment Adviser. Gaurav and their ‘relative(s)’ (as defined under Section 2(77) of the Companies Act, 2013) do not hold any financial interest in the companies mentioned in this article as of the date of publication. The views/recommendations mentioned in this article, wherever applicable, are those of the respective Sebi-registered Research Analyst/brokerage and have been reproduced/reported with due attribution. They should not be construed as the views or recommendations of The Economic Times Digital or the journalist. Readers are advised to consider the original research report and make their investment decisions based on their own assessment. Brokerage disclaimers here