The artificial intelligence trade is entering a new phase—one that extends far beyond chips. NVIDIA’s latest results revealed a staggering 263% surge in networking revenue, signaling that the real battleground in AI is no longer just about processing power, but the entire infrastructure stack that supports it. As demand for AI data centers accelerates, investors are increasingly shifting focus toward the broader ecosystem powering this transformation.

From GPUs to Full-Stack AI Infrastructure

Nvidia has long been the centerpiece of the AI boom, driven by explosive demand for its GPUs. However, the company’s networking business is now emerging as a critical growth engine. With quarterly networking revenue reaching $11 billion, technologies such as NVLink, Spectrum-X Ethernet, and InfiniBand are becoming indispensable for linking thousands of GPUs into unified systems.

This shift highlights a key evolution in AI infrastructure. Training and deploying advanced AI models require not only compute power but also ultra-fast data transfer and low-latency communication between chips. In this context, networking is no longer a supporting component—it is a core enabler of performance.

Despite already delivering a stock gain exceeding 1,000% since 2022, Nvidia continues to project massive future demand, with expectations of up to $1 trillion in cumulative orders for its next-generation chips through 2027. Yet, the company’s valuation—around 21 times forward earnings—suggests lingering skepticism about the sustainability of AI spending.

Rising Players Across the AI Supply Chain

The expansion of AI infrastructure is creating opportunities for companies beyond chipmakers. Arista Networks has emerged as a key beneficiary, leveraging its expertise in high-performance Ethernet switching. With revenue reaching $9 billion in 2025 and AI-related networking sales projected to double to $3.2 billion in 2026, Arista is positioning itself as a critical competitor—and complement—to Nvidia’s networking offerings.

Meanwhile, Vertiv Holdings is capitalizing on a different bottleneck: power and cooling. As AI data centers scale, the physical constraints of electricity and heat management are becoming increasingly important. Vertiv’s order growth of over 250% and a backlog of $15 billion underscore how infrastructure limitations are driving demand for specialized solutions.

Together, these companies illustrate a broader trend: the AI boom is no longer concentrated in a single segment. Instead, it is cascading across the entire value chain, from semiconductors to networking to physical infrastructure.

Opportunities and Risks in the Expanding AI Trade

For investors, the diversification of the AI trade presents both opportunities and challenges. On one hand, expanding exposure beyond GPUs allows participation in multiple layers of growth, potentially reducing reliance on a single dominant player. On the other hand, it introduces new risks tied to capital expenditure cycles and infrastructure spending.

A key variable to watch is the sustainability of data center investment. If hyperscalers and enterprises continue to ramp up spending, the entire ecosystem could benefit. However, any slowdown in capital expenditures would likely impact all segments simultaneously, from chips to networking to power systems.

Valuation is another consideration. While Nvidia appears relatively inexpensive given its growth, companies like Vertiv are trading at higher multiples, reflecting strong expectations that may already be priced in.

Looking ahead, the AI market is evolving into a full-scale infrastructure supercycle. The question is no longer whether AI will grow, but how broadly that growth will be distributed across industries. Investors who recognize this shift early may find opportunities beyond the obvious leaders—but success will depend on identifying which parts of the ecosystem can sustain long-term demand in an increasingly competitive landscape.