“The [AI] stories we are telling about $10 trillion, $15 trillion, $20 trillion, $25 trillion markets are actually terrifying stories for the rest of the world.” – Aswath Damodaran – Kerschner Family Chair in Finance Education, Professor of Finance at Stern School of Business of New York University

Multi-trillion narratives in technology markets usually begin as a convenience for investors and founders, but in artificial intelligence they have mutated into a systemic risk story for the global economy.¹ The promise of vast new revenue pools is being used to rationalise unprecedented capital expenditure, extreme valuations and business strategies that only make sense if global profit pools are effectively reshaped around a handful of platforms.^1,19,22 When those stories describe 10 trillion, 15 trillion, 20 trillion, 25 trillion markets, they are no longer just optimistic forecasts for a single sector; they imply a redistribution of value and power that other industries, workers and countries can only experience as loss.^1,10,16,22

Big market stories and the mechanics of delusion

Modern growth equity thrives on what Damodaran calls the “big market” story: the claim that the addressable market is so enormous that almost any current valuation can be justified because the upside is seemingly unbounded.^1,16,23 The conceptual tool doing the work here is total addressable market (TAM), defined not as the revenue a given firm can plausibly earn, but as the aggregate annual spend in a broad category that might be touched by the new technology.² In formal terms, if N is the number of potential customers and \bar{R} the average revenue per customer, the notional TAM is \text{TAM} = N \times \bar{R}.² The problem in speculative cycles is that both N and \bar{R} quietly expand over time as promoters redraw category boundaries and assume higher monetisation, turning a conventional planning metric into a story-telling device.

Past episodes in ride-hailing, streaming, cannabis and clean-tech all followed a similar script.^5,8,16,23 First, the existing market is measured narrowly (for instance, metered taxis) and appears too small to justify the valuations of the most ambitious firms.^5,8 Next, advocates argue that the new business model will not just capture that market but dramatically extend it by converting adjacent behaviours: private car use, public transport, or time spent on unrelated activities.^5,8,23 The TAM can then rise from 100 billion to 450 billion and on to 1 350 billion, allowing later investors to believe there is still plenty of headroom.^5,8 In AI, that same logic has been amplified: rather than counting only spend on software tools or cloud services, some prospectuses now define the potential market as all tasks that might conceivably be touched by AI, yielding figures in the tens of trillions.^1,11,16,22

This inflation is not mainly an exercise in arithmetic; it is a shift in narrative power. Once a market is framed as 20 trillion, any individual firm can present a path to multi-hundred-billion revenues without appearing absurd, and a host of smaller firms can style themselves as “picks and shovels” to the same supposed gold rush.^10,16,22 That is why Damodaran groups such episodes under the “big market delusion”: too many entrepreneurs, investors and incumbent CEOs simultaneously believe they can be dominant, discounting basic competitive dynamics and the inevitability of margin pressure.^1,4,16

The AI capital cycle and the new trillions

The current AI investment wave is unusually capital intensive for a software-led revolution.^1,7,19,22 Large language models and other frontier systems require vast data centre build-outs, high-end chips, energy capacity and network upgrades. Goldman Sachs has framed this as a 5,3 trillion capital spending cycle, mostly front-loaded into data centre infrastructure and semiconductor capacity.⁷ That spending, in turn, is justified to boards and investors using TAM slides showing 10 trillion to 25 trillion eventual revenue potential from AI-enabled productivity, automation and new digital services.^1,7,10,11,19

Damodaran’s concern is that there is now a visible gap between the aggregate market capitalisation and capex commitments of AI leaders on one side and any realistic path to revenues and profits on the other.^{1,10,16,19,22} In his own work, he estimates the overall AI market in high-end scenarios at 2 trillion to 3 trillion of annual revenues, already an extraordinary figure compared with today’s software and cloud markets but still far short of the most aggressive stories.¹⁶ When he computes the breakeven revenues needed for a single firm such as Nvidia to justify a 4,3 trillion market capitalisation, the required trajectory is stark: roughly 590 billion in annual revenues by 2030, implying growth of about 26% per year for the next five years.¹⁰ If several firms are priced as though they each can achieve such outcomes simultaneously, the gap between narrative and feasible demand becomes an issue not just for shareholders, but for the allocation of global savings.

In the AI context, the big market story carries an additional twist. Instead of simply promising to grow by outcompeting incumbents within a given sector, the leading platforms implicitly promise to boost productivity across most of the economy, thereby expanding global GDP and unlocking novel demand.^1,3,9,15,16 This allows promoters to argue that even if AI displaces incumbents and compresses margins in multiple industries, the net outcome will still be a larger pie, and therefore the multi-trillion TAM figures are self-consistent. Damodaran’s discomfort is not with the idea that AI might be transformative; it is with the leap from that prospect to the assumption that a handful of firms can sustainably extract and retain a double-digit share of the additional surplus.^1,10,16,22

Why multi-trillion AI TAMs imply loss for others

To understand why the headline numbers are “terrifying” for the rest of the world, it helps to unpack what a 20 trillion AI market would mean in flow terms.^1,10,16 Global GDP is currently in the order of 100 trillion to 110 trillion, depending on the measure.^3,9 Claiming that AI represents a 20 trillion revenue opportunity is roughly equivalent to asserting that close to one-fifth of total annual economic output can be intermediated or mediated by AI-centric platforms. Two mechanical consequences follow.

First, unless one assumes unprecedented GDP growth driven entirely by AI, much of that revenue must be reallocated from existing spend on labour, analogue processes and traditional vendors.^3,9,15,16 In other words, one sector’s TAM is another sector’s cost compression or outright obsolescence. If AI tools are to generate 20 trillion in revenues, either customers need to increase their total budgets dramatically or they must cut other line items. For non-tech sectors, this translates into a margin squeeze that could last for years, especially if AI providers initially use aggressive pricing to gain share and later raise prices once dependencies are locked in.^4,15,16

Secondly, even if AI-driven productivity improvements enlarge the total pie, the distributional pattern may be highly skewed. Platform economics tend to produce winner-takes-most outcomes, in which a few firms capture outsized profits while users, complements and competitors receive more modest gains.^3,9,16 If AI infrastructures are owned and controlled by a tiny group of US-based giants, the rent extraction from a 10 trillion to 25 trillion market will be concentrated geographically as well, exacerbating cross-border imbalances and reducing the policy autonomy of states that become dependent on foreign AI infrastructure.^3,9,22

From TAM to valuation: the translation problem

Damodaran’s professional lens is valuation, not macro commentary, and his scepticism emerges from the arithmetic of converting TAM narratives into discounted cash flows.^1,10,16,22 In a standard framework, the value of a firm is the present value of expected future free cash flows.^17,26 If FCF_t denotes free cash flow in period t and r the discount rate, the intrinsic value V is given by V = \sum_{t=1}^{\infty} \frac{FCF_t}{(1 + r)^t}.^17,26 A large TAM can justify high early growth rates in FCF_t, but only if the firm can convert revenue into cash at sustainable margins in the face of competition, regulation and technological change.^1,10,16

In AI, the chain from TAM to FCF_t is fragile at several links. Customer willingness to pay for AI outcomes remains uncertain; many pilots have yet to prove that they produce clear revenue gains or cost savings, and governance frameworks for measuring ROI are often immature.^6,15 Cost structures are also in flux: training and inference demands are rapidly increasing, energy prices are volatile, and model architectures may evolve in ways that undermine the scarcity value of today’s most expensive infrastructure.^1,7,10,16,19 Meanwhile, competitive dynamics are intense, with multiple hyperscalers and foundation model labs engaged in arms races on both capability and price.^1,3,9,16,19

Damodaran argues that when several firms are simultaneously valued as if each can secure a dominant share of a 10 trillion-plus market, investors are implicitly double-counting the same revenues.^1,16,22 Put differently, if two firms are each priced as though they will achieve 30% market share in a 10 trillion TAM, the combined implied market is 6 trillion even though the underlying category may realistically support barely half that amount. This type of inconsistency is a hallmark of big market delusions: investors aggregate optimistic assumptions firm by firm without reconciling them at the sector level.^1,16

Intangibles, narratives and the changing nature of value

Part of what makes AI so ripe for inflated TAM stories is the broader shift towards intangible-heavy business models. Modern technology firms derive much of their value from software, data, brand and organisational capital rather than physical assets.^22,23 These intangibles are scalable, making it plausible that a successful platform could reach extraordinary margins and global penetration once built. For AI, that logic is even stronger: once models exist, they can be fine-tuned and deployed across endless use cases at low marginal cost, or so the story goes.^1,3,16,19

Damodaran has spent years wrestling with how to value such intangible-intensive firms, and he acknowledges that traditional asset-based metrics understate their potential.^20,23 However, he also stresses that intangibles make narrative discipline more important, not less.^22,23 Because there are fewer hard anchors in plant, property and equipment, shifts in sentiment can swing valuations dramatically, and stories about future dominance can drift far from operational reality. Multi-trillion TAM slides, especially in AI, are therefore not simply descriptive; they are instruments for manufacturing belief in a future that may never arrive.^1,10,16,22

The tension is particularly acute when intangibles are leveraged to justify extreme capex. In many AI strategies, firms blend intangible assets (proprietary models, data, developer ecosystems) with tangible build-outs (chips, data centres, undersea cables). The more the narrative inflates future AI revenue pools, the easier it becomes for management teams to commit to massive, debt-funded capital programmes today.^1,7,16,19,22 If the narrative later deflates, the tangible commitments remain, and the adjustment falls on workers, suppliers and public finances as firms attempt to shed costs or seek policy support.

Strategic tension for incumbents and challengers

The multi-trillion AI market story creates a strategic trap for both established firms and high-growth challengers. On one side, incumbents in non-tech sectors fear being “left behind” and thus feel compelled to demonstrate AI activity to boards, investors and regulators.^6,15,21 That often translates into a proliferation of pilots, partnerships and capability-building programmes, with budgets justified partly by reference to sector-wide TAM forecasts. Yet without rigorous measurement of realised business value, organisations slip into what CIO advisers call the “activity trap”: plenty of motion, little confirmed impact.⁶

On the other side, start-ups and fast-growing technology firms face pressure from investors to present AI strategies that can plausibly scale to billions in revenue within a few years.^1,9,15,16 In pitch decks and S-1 filings, referencing industry TAM projections in the tens of trillions becomes almost obligatory, even if the company’s own serviceable obtainable market is a tiny fraction of that.^2,11,16 The more firms play this game simultaneously, the more crowded the supposed opportunity becomes, reinforcing Damodaran’s observation that not everyone can be right, yet few actors are willing to admit that they might belong to the long tail of marginal players.^1,4,16

At the sector level, these dynamics interact to produce what some analysts already describe as an emerging bubble in AI-related equities.^3,9,12,24 Valuations of the leading AI-exposed stocks have soared, private market pricing for model labs has reached extraordinary levels, and capital is chasing AI themes across semiconductors, infrastructure and application layers.^{1,3,9,12,19,24} Some commentators argue that the underlying technological advance is genuine and that current valuations are high but not yet irrational, pointing to robust earnings growth and near-term demand for compute.³ Others, including Damodaran, warn that the combination of extreme TAM claims, thin evidence of durable monetisation and aggressive capex has all the hallmarks of a classic boom that will eventually expose significant misallocation.^1,9,10,16

Debates and objections: could the trillions be real?

To be fair, there are thoughtful counterarguments. One line of reasoning holds that AI could trigger a new golden era of productivity, analogous to electrification or the spread of the internet, thereby supporting both substantial growth in global GDP and very large profit pools for technology providers.^3,9,15 If AI raises output per worker and enables novel products, total economic activity could expand enough that a 10 trillion-plus AI market does not require brutal cannibalisation of existing sectors. Moreover, proponents argue that AI is different from past tech cycles because it pervades almost every industry, from healthcare and finance to manufacturing and entertainment, justifying unusually broad TAM definitions.^3,9,15

A second objection is that early-return evidence is beginning to accumulate: hyperscalers are reporting strong demand for AI cloud services, chipmakers are enjoying spectacular revenue growth, and some enterprises are documenting tangible efficiency gains from AI automation.^1,3,9,12,16 Under this view, Damodaran may be underestimating the capacity of leading firms to defend high margins through vertical integration, ecosystem lock-in and rapid product cycles. If a small number of companies become indispensable providers of both infrastructure and application layers, they might indeed sustain multi-hundred-billion revenues each from AI-related services.^3,9

Damodaran’s response is not to deny these possibilities but to insist that valuation must reflect probabilities rather than stories alone.^1,10,16 Assigning a 100% probability to best-case TAM scenarios is analytically lazy. A more rigorous approach would model several paths for AI adoption, competition and regulation, discounting cash flows accordingly. The result is still likely to be a large prospective market, but it rarely stretches to the 20 trillion or 25 trillion headline figures that populate promotional materials.^10,11,16 For him, the risk is not that AI could end up smaller than hoped; it is that capital structures, public expectations and strategic commitments are already being set as though only the rosiest outcomes are acceptable.

Why the rest of the world should care

The most unsettling implication of the current AI TAM narrative lies outside the financial markets. When a small number of companies credibly promise investors that they can extract 10 trillion to 25 trillion of annual revenues from AI, they are implicitly making a claim about who will set the terms of digital production, communication and decision-making in the coming decades.^1,3,9,16 That claim has geopolitical, social and ethical dimensions.

Geopolitically, the consolidation of AI profit pools in a few jurisdictions could deepen the divide between technology exporters and technology importers.^3,9 Countries that lack domestic hyperscalers or foundational model champions may find themselves permanently on the buying side of a market whose rules they did not set, funnelling large streams of payments offshore in exchange for critical infrastructure and services. For them, the story of a 20 trillion AI market is not an exciting growth prospect but a future liability that could weaken their balance of payments and reduce their strategic autonomy.

Socially, a world in which AI platforms intermediate a large share of economic activity raises questions about labour displacement, bargaining power and privacy.^15,21 If AI tools are mainly deployed to reduce headcount and compress wages in non-tech sectors while generating rising rents for platform owners, income and wealth inequality is likely to widen.^3,9,15 The larger the TAM numbers become, the more they represent not just potential innovation but potential transfer from workers and smaller firms to a concentrated capital base. Civil societies and regulators therefore have a stake in scrutinising how AI value is created and how its gains are shared.

Ethically, the scale implied by 10 trillion to 25 trillion AI markets means that AI systems will influence an ever-expanding range of human choices, from credit allocation and medical diagnosis to hiring and political communication.^15,21 Concentrating that influence in a handful of profit-maximising entities, underpinned by valuations that assume relentless growth in monetisable engagement, creates obvious risks. It encourages designs and business models that prioritise data extraction, attention capture and automation of judgement, sometimes at the expense of transparency, autonomy and fairness.^15,21

Damodaran’s warning therefore operates on two levels. For investors, it is a call to resist the seduction of big market narratives that cannot be reconciled with grounded, probability-weighted cash flow projections.^1,10,16 For the broader public, it is a prompt to recognise that when AI stories scale to 10 trillion and beyond, they imply not just spectacular upside for a few firms but profound shifts in who pays, who benefits and who decides. Those stories are exhilarating for the companies at the centre of the AI boom; for the rest of the world, they may signal a coming era of dependence and disruption that deserves far more scrutiny than the TAM slides allow.^1,3,9,16,22

 

References

1. Aswath Damodaran on SpaceX, AI and the Big Market Delusion – 2026-06-18 – https://www.youtube.com/watch?v=vWx3kQuBHzE

2. Top Valuation Expert Says AI Market Needs ‘Trillions In Revenue’ To … – 2026-01-27 – https://finance.yahoo.com/news/top-valuation-expert-says-ai-033111714.html

3. Total Addressable Market – What Is TAM & How to Calculate It – Toptal – 2017-08-24 – https://www.toptal.com/management-consultants/market-sizing/total-addressable-market-example

4. Has the AI hype in equities gone too far? – VP Bank – 2025-11-27 – https://www.vpbank.com/research/en/content/59629/has-the-ai-hype-in-equities-gone-too-far

5. Wall Street’s valuation dean: AI has created a big market delusion – 2026-03-02 – https://www.thetimes.com/business/economics/article/ai-big-market-delusion-wall-street-valuation-dean-says-l96k86bgq

6. Gurley v. Damodaran on Uber’s Valuation – 2014-07-12 – http://www.allenlatta.com/allens-blog/gurley-v-damodaran-on-ubers-valuation

7. AI hype to AI value: Escaping the activity trap – CIO – 2026-04-22 – https://www.cio.com/article/4161509/ai-hype-to-ai-value-escaping-the-activity-trap.html

8. Aswath Damodaran Warns AI CapEx Boom Risks Debt-Driven … – 2026-06-19 – https://digg.com/tech/bi662l1z

9. How to Miss By a Mile: An Alternative Look at Uber’s Potential … – 2014-07-11 – https://abovethecrowd.com/2014/07/11/how-to-miss-by-a-mile-an-alternative-look-at-ubers-potential-market-size/

10. Valuing AI: Extreme Bubble, New Golden Era, or Both – GMO – 2026-01-13 – https://www.gmo.com/americas/research-library/valuing-ai-extreme-bubble-new-golden-era-or-both_viewpoints/

11. Trillion Dollar Market Caps: Fairy Tale Pricing or Business Marvels? – 2025-12-03 – https://aswathdamodaran.substack.com/p/trillion-dollar-market-caps-fairy

12. Full Transcript: Aswath Damodaran on Valuing SpaceX and AI – 2026-06-19 – https://excessreturnspod.substack.com/p/full-transcript-aswath-damodaran-cb4

13. AI-fueled market bubble concerns emerge as Big Tech … – YouTube – 2025-09-25 – https://www.youtube.com/watch?v=78f_lPJs4-A

14. 3 AI Stock Winners & 3 Write-Offs – Prof. Damodaran – YouTube – 2026-02-27 – https://www.youtube.com/watch?v=sXIU7V_PWX0

15. A “Fairly Highly Valued” Market: The Fed Chair Opines on Stocks … – 2025-10-06 – https://www.linkedin.com/pulse/fairly-highly-valued-market-fed-chair-opines-stocks-should-damodaran-irfgc

16. AI hype vs. reality: Assessing short-term buzz and long … – RSM US – 2024-09-10 – https://rsmus.com/insights/industries/technology-companies/ai-hype-versus-reality.html

17. [PDF] AI’S WINNERS, LOSERS AND WANNABES: BEYOND BUZZ WORDS! – https://pages.stern.nyu.edu/~adamodar/pdfiles/country/AIshort.pdf

18. [PDF] Estimating Risk Parameters Aswath Damodaran – NYU Stern – https://pages.stern.nyu.edu/~adamodar/pdfiles/papers/beta.pdf

19. Are the AI record valuations masking deeper problems in the market? – 2025-09-26 – https://www.reddit.com/r/investing/comments/1nqqobd/my_take_on_the_biggest_risk_are_the_ai_record/

20. Home Page for Aswath Damodaran – NYU Stern – https://pages.stern.nyu.edu/~adamodar/

21. The Dangers of AI Hype: How Misinformation Is Shaping Corporate … – 2026-02-22 – https://www.linkedin.com/pulse/dangers-ai-hype-how-misinformation-shaping-corporate-david-linthicum-ypxte

22. Aswath Damodaran on SpaceX, AI and the Big Market Delusion – 2026-06-19 – https://excessreturnspod.substack.com/p/the-trillion-dollar-gap-aswath-damodaran

23. Aswath Damodaran’s Blog, page 2 – Goodreads – 2026-02-01 – https://www.goodreads.com/author/show/3691.Aswath_Damodaran/blog?page=2

24. Are AI Bubble Concerns Warranted or Overblown? – YouTube – 2025-11-11 – https://www.youtube.com/watch?v=QldeBxgSLP0

25. Aswath Damodaran on SpaceX, AI and the Big Market Delusion – https://open.spotify.com/episode/1xrNJH7f3eCuRIwmmzIija

26. Equity Risk Premiums (ERP): Determinants, Estimates and … – 2026-03-11 – https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6361419