U.S. Stocks Are Overvalued, But Not In A Bubble | Professor Aswath Damodaran on Equity Valuations, AI Data Center Boom, and “Big Market Delusions”

Episode Summary

Executive Summary: Professor Aswath Damodaran argues that U.S. equities are richly priced but not obviously in a classic bubble, and that AI’s biggest mispricing is in private LLM companies rather than the chip/cloud infrastructure beneficiaries. He sees bubbles and corrections as normal features of innovation, while favoring disciplined valuation, cash flows, and selective ownership in mega-cap tech, gold as insurance, and caution toward overhyped private AI bets.

Main Topics: U.S. equity valuation and the risk of overpricing (Priority: 5/5): Damodaran says the market is richly priced and assumes a benign economic path, but he stops short of calling it a proven bubble because the numbers do not yet justify that label. AI as a ‘big market delusion’ and where the bubble is forming (Priority: 5/5): He distinguishes between AI infrastructure (chips, data centers, power) and LLM/application businesses, arguing the delusion is more likely in private model companies than in the architecture layer. Why public AI winners may still benefit even if the bubble bursts (Priority: 4/5): NVIDIA, Microsoft, and other hyperscalers may suffer slower growth if AI spending cools, but much of their revenue and cash generation is not solely dependent on AI, limiting systemic distress. Valuation philosophy: cash flows, buybacks, and earnings quality (Priority: 5/5): Damodaran emphasizes that valuation should be based on cash returns, especially dividends plus buybacks, not just earnings or simplistic PE ratios; he argues buybacks are the dominant return mechanism now. Platform ecosystems and network effects in tech (Priority: 4/5): He argues that durable value lies in ecosystems (Google/YouTube, Meta, cloud platforms) where products reinforce user lock-in and advertising or service monetization. Individual stock judgments: NVIDIA, Microsoft, Google, Meta, Oracle, Tesla (Priority: 5/5): He shares position-specific views: NVIDIA is richly priced and he would not buy today; Microsoft remains defensible in his portfolio; Google still leans heavily on ads; Meta’s ROI should be measured by time spent; Oracle looks overbet; Tesla is too political and story-driven for his taste. Gold as catastrophe insurance (Priority: 3/5): He frames gold as a hedge against loss of trust and tail risk, noting heightened concern about geopolitical and financial instability.

Key Arguments: The S&P 500 is richly priced because investors are assuming a relatively benign economic and geopolitical path, even though political, economic, and war-related shocks are not fully priced in. Market bubbles and corrections are a feature of innovation; overconfidence is necessary for major technological change, even though it creates overvaluation and eventual cleanup phases. The AI ‘bubble’ is more likely in private LLM companies than in infrastructure companies, because the infrastructure spend is real and observable while the end-product business model remains uncertain. Collectively, OpenAI, Anthropic, and xAI-like businesses look overvalued because their valuations imply revenue and profit scales that are not yet supported by the likely end state of the market. LLMs are not products by themselves; they are enabling infrastructure, and the real monetization will come from companies using them to create customer-facing products and services. NVIDIA, chipmakers, and data-center providers are relatively protected because their revenues are already earned; if AI demand slows, the damage is to future growth, not recovery of past cash spent. Microsoft and Meta are less exposed to an AI collapse than many assume because their core businesses (cloud, software, advertising) remain valuable and profitable independently of AI. Valuation should be based on cash flows returned to shareholders, especially dividends plus buybacks, because earnings alone are not cash flows and dividend-only models badly understate market cash generation. AI may reduce costs for individual firms, but if everyone gets the same tool, competition pushes prices down and margins can fall rather than rise across the economy. Tesla is valued on a narrative of future businesses like robotics and autonomous driving, not on cars; Damodaran is unwilling to pay for that uncertainty and dislikes the political overlay. Gold is a rational insurance asset when trust in systems and institutions weakens. Even if AI investments are impaired, the largest public tech firms likely avoid bankruptcy because their cash flows are strong enough to absorb mistakes.

Data Points: S&P 500 valuation: Richly priced / overvalued - Damodaran says the market is priced for a benign economic outcome and remains expensive by his framework. Equity risk premium (January estimate): 4.23% - He says this is around the median of the last 60 years, low versus post-2008 but not extreme historically. OpenAI valuation rumor: $500B current, rumored up to $850B-$1T - Used to illustrate how stretched private AI valuations have become. OpenAI annual revenue run rate: $20B - Referenced against its implied valuation to show a very high sales multiple. OpenAI estimated 2026 revenue: $30B - Used to frame a forward price-to-sales multiple of roughly 30x or more. Private AI valuation multiple: ~30x forward sales - Damodaran cites this as evidence that LLM companies are collectively overvalued. Microsoft ownership period: Owned since 2014 - He says he still holds it fully in his portfolio. NVIDIA ownership period: Since 2018 - He notes he sold his last quarter at the end of last year in staged sales. NVIDIA forward P/E: ~30x - Mentioned as a rough estimate for how richly priced it is. NVIDIA trailing P/E: ~50x - Used to argue that too much growth is already embedded in the price. Meta user time in ecosystem: 57 minutes per day - He suggests this is a key metric for measuring AI-driven engagement and ad potential. Alphabet/Google corporate structure: "Alphabet is really six to one and one giant" - Used to argue many moonshot bets have not become standalone value creators. U.S. corporate cash returns: ~85% of earnings returned in cash - He says buybacks plus dividends explain why equity risk premiums remain reasonable. Buybacks last year: More than $1 trillion - Cited to show how shareholder returns are supported beyond dividends. Dividend yield: ~1.5% to 2% - Too low on its own to justify valuation, in his view. Cash yield including buybacks: ~4% - He argues this is the appropriate cash-return metric for valuation. S&P 500 tech weight: ~30% of market cap - Used to explain why earnings have stayed resilient and why tech drives index-level stability. Tesla forward P/S: 15x - He says Tesla’s valuation requires a shift into a higher-margin business. Tesla forward P/E: 220x - Used to show that auto-only fundamentals cannot justify the stock.

Pivotal Quotes: "“It’s a richly priced market which is building in expectations that the pathway is going to be a benign one.”" — Professor Oswath DeModarin: Opening assessment of the S&P 500 and broader market valuation. "“There is a big market delusion. There will be a correction as a consequence, but I think that’s part of being in a market.”" — Professor Oswath DeModarin: His core thesis on bubbles as a recurring feature of innovation and human overreach. "“LLMs are part of the architecture. They’re not a product to service by themselves.”" — Professor Oswath DeModarin: Explaining why he sees private model companies as the weakest link in the AI value chain.

Implications: Listeners should distinguish between AI infrastructure winners and speculative LLM bets, and use cash-flow-based valuation rather than hype or PE-only narratives. The big risk is a cleanup phase in private AI, not necessarily a collapse of mega-cap tech.