In the Loop – July 10, 2026

One of the most important developments in today’s market is that artificial intelligence is no longer just a software story, it has become a full-scale industrial investment cycle. Just a few years ago, hyperscale cloud providers collectively spent roughly $162 billion on capital expenditures. Today, annual spending is on pace to approach $600-$750 billion in 2026, with many industry forecasts suggesting investment could move even higher over the next several years. This is no longer simply a technology cycle; it is a generational buildout of physical infrastructure requiring enormous investments in semiconductors, networking, power generation, electrical equipment, cooling systems, construction, and data centers. In our view, investors should increasingly think of AI as an infrastructure theme rather than a software theme.

The drivers behind this spending continue to strengthen. AI adoption is expanding across both consumers and enterprises, while reasoning models and autonomous AI agents require significantly more computing power than earlier generations of large language models. At the same time, token usage continues to accelerate at an extraordinary pace as AI becomes integrated into everyday workflows. Each new generation of models demands more inference-time compute, creating a virtuous cycle where higher usage leads to greater infrastructure investment, which in turn enables even more powerful AI capabilities.

This week’s earnings from the memory industry reinforced that trend. Micron Technology delivered another strong quarter, highlighting continued demand for high-bandwidth memory (HBM) and AI-related DRAM despite investors taking profits following the stock’s significant run over the past year. Samsung Electronics also reported results that reflected ongoing strength in AI memory demand, although investors remained focused on execution and competitive positioning within HBM rather than near-term earnings alone. The market’s reaction illustrates that expectations have become exceptionally high, but importantly, neither company suggested that AI infrastructure demand is slowing. Instead, both reinforced the view that memory remains one of the primary bottlenecks supporting the AI investment cycle.

Another notable development came from SK Hynix, which completed the largest-ever foreign ADR listing in U.S. history, raising approximately $26.5 billion. The decision to access U.S. capital markets reflects both the enormous capital requirements of the AI memory industry and management’s desire to broaden its shareholder base while benefiting from the premium valuations often afforded to leading U.S.-listed semiconductor companies. The offering also gives U.S. investors direct exposure to one of the world’s leaders in high-bandwidth memory, an area that remains supply constrained as hyperscalers continue expanding AI infrastructure.

For investors, the most important question is no longer simply which AI model will win, but rather where every dollar of AI capital spending ultimately flows. Every new AI data center requires advanced chips, HBM memory, advanced packaging, servers, networking equipment, power distribution, cooling systems, and reliable electricity before a single AI query is ever processed. Those physical bottlenecks remain where we believe the most durable long-term investment opportunities exist.

This continues to reinforce our U.S. Resiliency investment framework. Rather than concentrating solely on AI software companies, we remain focused on the mission-critical businesses enabling this unprecedented infrastructure buildout—companies involved in electricity generation, grid modernization, networking, advanced semiconductors, industrial construction, and critical physical infrastructure. As AI transitions from a technology theme into an industrial economy theme, we believe owning the companies that build the backbone of this ecosystem offers a compelling way to participate in what may become one of the largest capital investment cycles in modern history.