A value investor’s framework for AI
If you haven’t noticed already, allow me to point out that investment discussions about artificial intelligence (AI) frequently devolve into almost insulting debates between AI evangelists, who cite massive total addressable markets (TAM) and explosive user growth, and sceptics, pointing to trillions of dollars in capital expenditure (capex), a buildout based on speculation, exorbitant compute costs, and an unclear path to profitability.
Aswath Damodaran, NYU Stern professor and widely respected value investor, believes this debate has largely gone off the rails, noting in a recent post, “The debate about AI, in my view, has gone off the track with advocates and sceptics often talking past each other”. Optimists cherry-pick usage metrics, while pessimists focus on heavy spending, with both sides ignoring the rigorous business analysis that needs to take place in the middle.
To move past the binary “Cultists vs. Luddites” rhetoric, Damodaran reinforces the need to evaluate AI based on unit economics, sustainable profit margins, and defensible moats.
The four phases of revolutionary change
Every game-changing General Purpose Technology (GPT) has followed a distinct hype cycle. Damodaran posits they have also followed a distinct life cycle and structures his version of the journey into four core stages:
- Hope and hype: Visionaries sell the narrative of transformational change to secure talent and funding. Because tangible products or earnings do not yet exist, narrative drives valuation.
- The investing build-up: Companies rush to spend the funds and construct the infrastructure required for the shift. This stage frequently triggers the ‘Big Market Delusion,’ in which overconfident entrepreneurs and venture capitalists collectively overinvest and overprice firms based on overlapping and double-counted market-share assumptions.
- Business building: Investors transition from pricing promises to demanding execution. Companies face the “bar mitzvah moment,” where they must establish working supply chains, unit economics, and pricing models that demonstrate a path, ultimately to profitability.
- Recalibration: Fundamental market forces take hold. Winners consolidate market share, while inefficient or obsolete incumbent businesses face Joseph Schumpeter’s creative destruction.
According to Damodaran, AI has built “the most expensive factory in history, and done so in hyper speed”. Hyperscalers like Microsoft, Meta, Alphabet, Amazon, Oracle, and CoreWeave have collectively committed over US$1.73 trillion in capital expenditures to build data centres and compute capacity. Yet, the industry is only now entering its “bar mitzvah moment”, shifting from infrastructure build-out to monetisation.
Framing the total addressable market (TAM)
Wall Street bankers frequently justify lofty valuations by citing astronomical addressable markets. For example, bankers estimated a US$22 trillion TAM for xAI – a figure Damodaran rightly calls “hallucination”.
To develop a realistic TAM model, Damodaran grounds the upper limit in corporate operating expenses. Worldwide, publicly traded companies spent approximately US$65 trillion on operating costs in 2025, while total employee compensation in the U.S. alone was about US$13 trillion. Since AI cannot replace physical raw materials such as rubber or wheat, its potential market mainly focuses on labour-related expenditures.
Damodaran then offers four critical variables that dictate where the actual market size lands:
- Tool vs. worker replacement: If AI functions strictly as a tool, corporate budgets for it will remain a small fraction of labour costs. If AI acts as an autonomous employee replacement, the potential market expands significantly.
- Income tiers: Premium AI agents (e.g., advanced coding models) carry higher development costs and target high-income knowledge workers, who account for roughly half of all U.S. compensation.
- Industry exposure: Sectors such as software development and financial services (which account for under 20 per cent of global operating expenses) are far more susceptible to immediate AI integration than capital-heavy sectors like industrials, utilities, or materials.
- Geography: High-income, service-heavy economies – primarily the U.S., Europe, and China – represent the dominant market opportunity.
Depending on where these variables settle, Damodaran estimates realistic TAM scenarios ranging from under US$1 trillion (basic productivity tools in select industries) to over US$10 trillion (broad worker replacement globally).
Industry economics: Unit costs and business models
Unlike classic software or platform businesses, where the marginal cost of serving an additional user approaches zero, AI faces severe unit-economic constraints.
Every query or token generated incurs direct compute and electricity expenses. As a result, early subscription-based models proved unprofitable when users ran heavy workloads. The market is consequently pivoting toward usage-based pricing models, particularly for high-end enterprise applications – this, of course, could limit the rate and ultimate scale of the uptake.
In mass-market AI, token generation costs are falling rapidly due to infrastructure efficiencies. However, in frontier/premium AI, scaling model power requires exponentially more data, energy, and advanced chips, keeping unit costs stubbornly high.
Importantly, competitive moats will differ by segment:
- Mass-market AI: Moats will rely primarily on economies of scale and low-cost token production.
- Premium enterprise AI: Moats will stem from sticky integrations with proprietary client data, deep technical know-how, and strict data privacy compliance.
Social constraints and valuations
It’s worth remembering value creation doesn’t happen in a vacuum. Damodaran notes that societal pushback can and will directly influence corporate cash flows and valuations through several bottlenecks.
For example, data centres command large land, water, and electricity footprints, which will trigger local political pushback and increased power utility rates.
Meanwhile, we will inevitably see data breaches and misuse scandals. These, in turn, will lead to tighter regulatory oversight and increased compliance costs.
And let’s not forget that if white-collar jobs are meaningfully displaced, the result will be a push for mandatory severance protections or policy interventions to protect employment.
Reverse engineering reasonable expectations
To apply this framework as an investor, Damodaran advocates using reverse-engineered breakeven analysis rather than relying on ungrounded growth assumptions.
For instance, if an AI frontier company targets an initial public offering (IPO) valuation of US$2 trillion, assuming a 30 per cent after-tax operating margin and a 10 per cent cost of capital, that single company would need to generate roughly US$1.2 trillion in annual revenue within 10 years to justify its pricing.
Presumably, Damodaran arrives at this conclusion by also implying a circa 14.5x after-tax operating earnings multiple in Year 10.
At an aggregate industry market capitalisation of US$5 trillion, the market is pricing in an end-state AI product and service revenue of US$5 trillion to US$8 trillion.
Comparing these implied requirements against current AI product revenues – which sit under US$250 billion – reveals a significant gap between market pricing and reality. That’s Damodaran’s diplomatic way (using a very conservative 10 per cent required return and or a 14x multiple) of saying there’s some hype in today’s market prices for these businesses.
Conclusion
As Damodaran observes, “Big markets don’t always become big businesses”. For investors trying to make sensible investment decisions amid the AI wave, you cannot be blindly enthusiastic nor cynical. But you do need to measure the hype against some reasonable assumptions about numbers.
Roger Montgomery is the Founder and Chairman of Montgomery Investment Management. Roger has over three decades of experience in funds management and related activities, including equities analysis, equity and derivatives strategy, trading and stockbroking. Prior to establishing Montgomery, Roger held positions at Ord Minnett Jardine Fleming, BT (Australia) Limited and Merrill Lynch.
He is also author of best-selling investment guide-book for the stock market, Value.able – how to value the best stocks and buy them for less than they are worth.
Roger appears regularly on television and radio, and in the press, including ABC radio and TV, The Australian and Ausbiz. View upcoming media appearances.
This post was contributed by a representative of Montgomery Investment Management Pty Limited (AFSL No. 354564). The principal purpose of this post is to provide factual information and not provide financial product advice. Additionally, the information provided is not intended to provide any recommendation or opinion about any financial product. Any commentary and statements of opinion however may contain general advice only that is prepared without taking into account your personal objectives, financial circumstances or needs. Because of this, before acting on any of the information provided, you should always consider its appropriateness in light of your personal objectives, financial circumstances and needs and should consider seeking independent advice from a financial advisor if necessary before making any decisions. This post specifically excludes personal advice.
