Speculative manias, like the AI boom, function like collective hallucinations. The overwhelming belief in future demand becomes self-fulfilling, attracting capital that builds tangible infrastructure (e.g., data centers, fiber optic cables) long before cash flows appear, often leaving lasting value even after the bubble bursts.

While there's a popular narrative about a US manufacturing resurgence, the massive capital spending on AI contradicts it. By consuming a huge portion of available capital and accounting for half of GDP growth, the AI boom drives up the cost of capital for all non-AI sectors, making it harder for manufacturing and other startups to get funded.

The previous era of central bank money printing lifted all asset classes together. The new regime, driven by private borrowing for real economic investment, is different. It creates GDP growth (good for stocks) but also a large supply of debt (bad for bonds).

Major investment cycles like railroads and the internet didn't cause credit weakness because the technology failed, but because capacity was built far ahead of demand. This overbuilding crushed investment returns. The current AI cycle is different because strong, underlying demand is so far keeping pace with new capacity.

Veteran investor Jim Schaefer notes a recurring pattern before recessions: a massive, euphoric movement of capital into a specific area (e.g., telecom in 2001, mortgages in 2008). This over-investment inevitably creates systemic problems. Investors should be wary of any asset class currently experiencing such a large-scale influx.

For 2026, AI's primary economic effect is fueling demand through massive investment in infrastructure like data centers. The widely expected productivity gains that would lower inflation (the supply-side effect) won't materialize for a few years, creating a short-term inflationary pressure from heightened business spending.

While low rates and high nominal growth typically favor equities, financial repression introduces a counterintuitive risk. If institutions are forced to buy government bonds, they must sell liquid assets—primarily equities. This could lead to a slow, multi-year decline in the S&P 500, mirroring the 1966-1982 period, instead of a sudden crash.

Asset allocation should be based on liquidity cycles, not economic cycles like GDP growth, as they are out of sync. An increase in liquidity precedes economic acceleration by 12-15 months. Strong economic data can even be a negative signal for asset markets as it means money is leaving financials for the real economy.

An anticipated $3 trillion in AI-related spending requires significant debt financing, creating a $1.5 trillion gap. This is expected to cause a 60% increase in net investment-grade bond issuance, creating a supply-side headwind that makes the asset class less attractive despite sound fundamentals.