Equities & sectorsOaktree Capital
A Layman Thinks About AI
Artificial intelligenceAI infrastructure
Our summary
AI’s transformative potential is matched by an unusually high level of uncertainty, so investors should assess the probability of being right across several linked forecasts before committing capital.
- Marks illustrates the problem with four forecasts, each assumed to have a 60% success rate: the probability that all are correct is only 13% if the forecasts are independent.
- AI development requires judgments about technology, computing demand, competition, user benefits, financing, timing, reliability, ethics, creativity, reasoning and the possible emergence of artificial general intelligence.
- The economic value may accrue to users rather than providers if comparable AI models become price-driven commodities, while dominant companies could capture large profits through essential products and pricing power. Nvidia and TSMC are cited as current examples of strong competitive advantages.
- Competitive leadership may change rapidly. OpenAI and Anthropic were almost unknown a few years ago, while Microsoft, Amazon, Apple and Google survived earlier technology shifts that displaced Napster, Yahoo and Netscape.
- Forecasting AI company profits through 2030 requires assumptions about unit sales, prices, costs, reinvestment, chip lives, data-centre depreciation and the duration of capital-spending competition.
Takeaway: AI merits investor attention because its potential is large, while the uncertainty around technology, market structure, taxation, profitability and leadership makes precise valuation especially fragile.
Published by Oaktree Capital. Summary by HF Best Ideas.