Dot-com peak was 25× price-to-revenue. AI is at 50×. The S&P 500 trades at 22× P/E. Either AI revenue must grow 5× in 3 years, or valuations must correct dramatically.
50×AI sector price-to-revenue multiple (2024)
$5TCombined AI company valuations (public + private)
$100BCombined AI company revenue (2024)
25×Dot-com peak multiple (NASDAQ 2000)
Choose your depth. The data doesn't change — just the explanation.
AI companies are worth $5 trillion total, but they only make $100 billion a year in revenue. That means people are paying 50 times more for these companies than what they actually make. During the dot-com bubble in 2000, the number was only 25 times. Normal companies trade at about 22 times their profits. AI is being valued at a level that requires enormous future growth to justify.
The 50× price-to-revenue ratio for AI companies means investors are betting the revenue will grow dramatically. The math: to justify a 50× multiple at normal (5× P/S) levels, revenue must grow 10×. At $100B in 2024, that means $1T in AI revenue by 2027-2028. McKinsey estimated $4.4T annual potential — but that's full deployment, which their own research shows only 8% of enterprises have achieved. The gap between what investors expect and what deployment rates support is the bubble's defining characteristic.
AI company valuation methodology: NVIDIA market cap $3.2T (P/S ~30× on $108B rev, but GPU sales tied to AI spending). OpenAI $157B implied valuation ($3.7B ARR = 42× P/S). Anthropic $60B implied ($1.3B ARR = 46× P/S). NASDAQ 2000 peak: price-to-revenue ~25× (Nasdaq composite). S&P 500 current P/E: ~22× (EPS-based). Historical tech multiple compression: when NASDAQ corrected 78% (2000-2002), revenue-multiple compressed from 25× to 4×. If AI compresses similarly, sector valuation falls to $400B — from $5T. Bull case: AI genuinely delivers $4.4T in productivity, justifying valuations retrospectively (similar to how Amazon's 1999 valuations were retrospectively justified by 2010).
Valuation data: Bloomberg Intelligence AI Sector Monitor, PitchBook, Crunchbase unicorn tracker. NVIDIA: finance.yahoo.com/quote/NVDA. OpenAI: The Information revenue projections (theinformation.com/articles/openai-revenue-and-loss-projections). Anthropic: TechCrunch, WSJ deal coverage. NASDAQ historical: finance.yahoo.com/quote/%5EIXIC/history (COMP index). P/S comparison: simply take market cap / annual revenue — different from P/E (uses earnings). Track: Epoch AI "Compute and Revenue Correlations," Goldman Sachs "AI Spending Review Q4 2024." Compare: hyperscaler AI capex trends vs. return-on-AI-investment reports from Brookings.
Historical Bubble Comparison
How does the AI valuation multiple compare to historical bubbles? The data suggests AI is more extreme than dot-com at its peak.
Price-to-Revenue Multiples: AI vs. Historical Bubbles
Comparison of sector-level price-to-revenue at peak and current levels.
Revenue Required to Justify Current AI Valuations
Revenue needed at various multiples to justify $5T AI sector value.
AI Revenue Projections — Current vs. Required
Actual trajectory vs. what valuation implies must happen.
The 5× Revenue Problem
For AI valuations to be "rational" at a 10× revenue multiple (respectable SaaS level), the sector must grow from $100B to $500B in revenue. In 3 years. That requires the enterprise adoption gap to close from 8% at scale to 50%+ while keeping pricing stable as APIs deflate. Possible? Yes. Certain? Not remotely.
Sources
• Bloomberg Intelligence. AI Sector Monitor. bloomberg.com/professional/solution/bloomberg-intelligence.
• The Information. "OpenAI Revenue and Loss Projections." theinformation.com. 2024.