AI Investment Thesis Shifts Toward Inference Amid Macro Headwinds

The initial surge in artificial intelligence equities has encountered significant resistance in 2026, with former leaders like Nvidia and Microsoft facing valuation pressures. While Nvidia has seen a modest 8% gain this year, Microsoft has experienced a sharp decline of nearly 60%, driven by persistent inflation and the Federal Reserve's reluctance to lower interest rates. This cooling period is attributed to a combination of macroeconomic instability and geopolitical tensions in the Middle East, which have pushed investors toward more conservative asset classes. However, the underlying secular trend of AI adoption remains intact, shifting from the 'training' phase to the 'inference' phase. The training market, dominated by large language models from OpenAI, Meta, and Alphabet, is characterized by cyclical and lumpy capital expenditures. In contrast, the inference market—the actual application of AI to generate outputs—is becoming a primary driver of spending. This shift benefits companies providing custom ASICs and CPUs, such as Broadcom and Intel. Beyond semiconductors, the expansion of data center infrastructure is creating demand for optical hardware from providers like Corning and Lumentum, as well as increased leasing for REITs such as Digital Realty. Energy requirements are also driving significant deals with providers like Brookfield Renewable. With a projected compound annual growth rate (CAGR) of 30.6% through 2033, the AI market is expected to expand as enterprises integrate agentic AI solutions. Some analysts argue that a potential recession could actually accelerate this trend as firms seek AI-driven efficiency to streamline operations and replace human labor.