As we move through the second quarter of 2026, the Artificial Intelligence (AI) investment landscape is undergoing a fundamental shift. For years, Nvidia was the undisputed hegemon, the gateway through which every dollar invested in AI had to pass. However, recent performance data from May 2026 reveals a new hierarchy. While Nvidia remains a highly profitable powerhouse, its stock appreciation has begun to level off, giving way to companies solving the most pressing problems of the current phase: energy efficiency and semiconductor specialization.
The Shift from Training to Inference
2026 will go down in history as the year of the "Great Shift to Inference." If 2023 and 2024 were about training giant models (LLMs), 2026 is about running those models at scale. This transition has diminished the relative advantage of Nvidia's general-purpose GPUs in favor of Application-Specific Integrated Circuits (ASICs). Companies like Broadcom and Marvell Technology have seen their stocks surge by 67% since the start of the year, as cloud giants (Microsoft, Amazon, Google) pivot to their own internally designed chips to reduce operational costs.
Broadcom, in particular, has emerged as the central pillar of AI infrastructure in 2026. With its dominance in networking technologies and its close partnership with Alphabet for TPU processors, the company is reaping the rewards of a market that now demands data transfer speeds as much as raw processing power. Data analysis shows that demand for custom AI silicon is growing at twice the rate of the general GPU market, explaining the outperformance against the Nasdaq index.
The Energy Wall and the Infrastructure Boom
Perhaps the most striking story of 2026 is the rise of companies managing the "physical" layer of AI. Vertiv Holdings, a firm specializing in cooling and power management for data centers, has recorded gains of 121% this year. The argument is simple but relentless: chips from Nvidia and its competitors produce so much heat and consume so much power that traditional data centers are unable to support them.
The adoption of liquid cooling became the industrial standard in early 2026, and Vertiv holds the largest market share in this sector. Investors are realizing that AI is not just code and semiconductors, but also copper, electricity, and water. The energy crisis triggered by the expansion of AI clusters has also led to a renaissance in nuclear energy and smart grid stocks, which are now treated as "indirect AI plays" with massive upside potential.
The Law of Large Numbers Hits Nvidia
Why is Nvidia lagging behind? The answer lies in mathematics. With a market capitalization that has reached stratospheric levels, Nvidia now requires massive capital inflows to achieve even a 10% gain. In contrast, mid-to-large cap companies that are part of the supply chain have a much longer "runway" for growth. Furthermore, competition from AMD and Intel, which in 2026 finally managed to offer viable alternatives for inference, has begun to squeeze Jensen Huang’s profit margins.
"The market is no longer looking for who will build the next big model, but for who will manage to keep it running without collapsing the power grid," says a leading Wall Street analyst.
In conclusion, 2026 marks the maturation of the AI market. Investors who focused exclusively on processor manufacturers missed the biggest opportunities of the year, which were hidden in infrastructure and specialized hardware. This trend is expected to continue as AI demand is now integrated into every aspect of the global economy, from automotive to heavy industry, requiring solutions that are efficient, not just powerful.
