Introduction

Artificial intelligence has moved from experimental labs into the core of daily business operations, driving a surge in compute demand, infrastructure spending, and new investment opportunities. As more firms integrate AI into routine workflows, the physical infrastructure required to support that shift is expanding at an unprecedented pace.

What Happened

Recent analysis shows that nearly half of all businesses could be running AI daily within six months, according to Goldman Sachs, which projects U.S. data center power demand will double to 66 gigawatts by 2027. NVIDIA reported Q2 revenue exceeding $96 billion, with data center sales jumping 117% to $89 billion, while Broadcom forecast AI semiconductor revenue reaching $230 billion by fiscal 2028. Vertiv's free cash flow surged 234% year-to-date, and Nebius Group revenue exploded 454% to $582 million, driven by strong AI cloud growth. These five companies—NVIDIA, Broadcom, Arista Networks, Vertiv, and Nebius—are positioned at the heart of the infrastructure buildout.

Why This Matters

The through-line is clear: top hyperscalers are on track to spend nearly $800 billion in capital expenditures in 2026 and $1.3 trillion in 2027, creating intense demand for compute, custom silicon, networking gear, power systems, and cooling solutions. With Goldman estimating U.S. data center power needs climbing from about 31 gigawatts in 2025 to 66 gigawatts by 2027, and the International Energy Agency projecting global data center electricity use to nearly double by 2030, the companies supplying the picks, shovels, wires, and power gear are seeing heightened investor interest. The shift also means that everyday business operations will increasingly depend on stable, scalable infrastructure that these five firms help provide.

Key Takeaways

  • NVIDIA remains the foundational compute layer, with Q2 data center revenue up 117% and a product roadmap that includes Vera Rubin, expected to deliver significant revenue per gigawatt.
  • Broadcom is capturing the custom silicon and AI networking market, projecting $230 billion in AI semiconductor revenue by fiscal 2028, powered by TPU deployments and its new 200-terabit Ethernet switch.
  • Arista Networks owns the Ethernet fabric that links thousands of GPUs into a single system, reporting strong quarterly growth and expanding its AI-fabric customer base beyond 100 enterprises.
  • Vertiv provides the power and thermal management systems AI data centers require, with free cash flow up 234% and co-development of 800-volt DC architectures with NVIDIA.
  • Nebius Group operates a neocloud bridging GPU scarcity and enterprise demand, with revenue jumping 454% and a contracted power target expanded to 5 gigawatts.

While each company offers significant upside, investors should weigh risks including memory pricing volatility, customer concentration, heavy capital expenditure, and dependence on a small group of frontier-model clients.

Conclusion

The AI infrastructure buildout is no longer a niche theme—it is reshaping the foundation of how businesses operate and how capital is allocated. With top-line growth driven by real demand curves and massive capex commitments, the five profiled names represent distinct exposure points to the surge, but also concentrated risk. As everyday AI adoption follows the adoption trajectory identified by industry research, the physical bottlenecks of compute, power, and networking will remain key determinants of which companies capture the most value in the years ahead.