AI Supply Chain Explained: Chips, Data Centers and Software, With the Bull and Bear Cases

Gartner forecasts worldwide AI spending of $2.67T in 2026, up 49.5% from about $1.79T in 2025, and most of it goes to infrastructure. Hyperscaler capex guidance of roughly $700–745B, Nvidia's $96.2B quarter and TSMC's revenue growth show the money reaching chips and foundries. The bear case rests on cyclical, price-led memory growth, heavy spending against rising depreciation, customer concentration, and a gap between infrastructure outlays and spending on AI models and agents. Thailand's link is mainly data center investment approvals, but our notes contain no Thai fund or SET data. This explainer sets out the value chain, general exposure routes and their risks, and what to monitor, without recommending anything.