In 2026, the AI investment narrative will expand from foundational model creators to companies building applications and services. It also includes sectors enabling AI growth, such as energy generation and data centers, offering a wider range of investment opportunities beyond the initial tech giants.
With inflation becoming less of a concern in 2026, bond yields will be driven more by growth expectations than inflation risk. This restores their traditional negative correlation with equities, making them a more reliable diversifier and hedge against a potential economic downturn in portfolios with long-risk exposure.
A new structural driver for gold is demand from emerging market central banks seeking to mitigate geopolitical risks. Events like the freezing of Russia's reserves have accelerated a trend of buying gold to reduce exposure to sanctions and to back their own currencies, creating a higher floor for prices.
Investors' equity allocations are high, not necessarily from new purchases, but from strong market performance. This passive 'drift' creates a significant, often overlooked, concentration risk. This means many portfolios are more exposed to an equity drawdown than their owners may realize, necessitating a review of diversification strategies.
A key risk for 2026 is the disconnect between stretched market valuations (e.g., tight credit spreads in the 1st percentile) and a macroeconomic environment that doesn't feel late-cycle. This tension suggests that even if growth drives equities higher, it could be accompanied by increased volatility or widening credit spreads.
With credit spreads already tight, their potential upside is limited while their downside is significant in a recession scare, offering poor convexity. Goldman Sachs advises that a better late-cycle strategy is to move up the risk curve via equities, which offer more upside potential, rather than through credit investments.
The massive, direct, and geographically concentrated energy demand from AI data centers makes local U.S. power markets the most effective AI-related commodity trade. With 72% of data centers in just 1% of counties and a constrained grid, local power prices are poised to rise significantly, offering a targeted investment thesis.