The artificial intelligence boom has created a familiar investing playbook: Buy the companies making the GPUs. That strategy has worked remarkably well, but it overlooks an increasingly important reality. A GPU is useless without a building to house it, electricity to power it, cooling to keep it running, and a grid connection capable of supplying the enormous load. The physical infrastructure supporting AI may ultimately prove just as important as the chips doing the computing.
That creates a different group of winners. Eaton (NYSE:ETN | ETN Price Prediction), Vertiv (NYSE:VRT), and Quanta Services (NYSE:PWR) sit at different points along that infrastructure chain. They are the picks-and-shovels companies supplying the electrical systems, cooling equipment and construction expertise required to turn AI spending into actual data centers.
Eaton Has a 15-Year Pipeline
Eaton may be the clearest example of why investors should look beyond GPUs. The company sells switchgear, power distribution equipment, uninterruptible power systems and increasingly sophisticated cooling infrastructure — the equipment that allows a data center to turn electricity into computing capacity.
During its latest earnings update, Eaton management said its U.S. data center backlog had reached 307 gigawatts (GW), equivalent to 15 years of work at 2025 build rates, up from 12 years previously. Only about 20% of that backlog is expected to convert in the near term, with most deliveries extending into 2028 and beyond.
That distinction matters. The 307-GW figure is not 307 GW of data centers under construction. Industry trackers have reported pipelines in roughly the 250-GW to 330-plus-GW range, but projects still face years-long grid interconnection queues, permitting, labor shortages, equipment lead times and financing hurdles.
For Eaton, however, that creates something investors value: visibility.
Vertiv Sits Inside the Data Hall
Moving closer to the GPUs, Vertiv is arguably the most direct pure-play among the three, supplying critical power and thermal-management equipment inside data centers.
Its second-quarter 2026 revenue rose 24% to $3.27 billion, while adjusted EPS increased 60% to $1.52. Vertiv also generated $925 million of adjusted free cash flow during the quarter. Management raised its full-year revenue forecast to $13.8 billion to $14.2 billion and expects adjusted earnings of $6.65 to $6.75 per share.
The investment thesis is straightforward: As AI racks consume more electricity, power distribution and cooling become harder problems. Liquid cooling, higher-capacity UPS systems, and advanced thermal management aren’t optional accessories. They’re part of the computing infrastructure.
Quanta Gets the Power There
Quanta Services plays a different role. It is closer to the construction crew than the equipment manufacturer, building transmission lines, substations and electrical infrastructure that connect massive new loads to the grid.
Its first-quarter 2026 backlog reached $48.47 billion, up from $43.98 billion at the end of 2025, while revenue climbed 26% to $7.87 billion. Its electric-infrastructure backlog alone reached $40.1 billion.
That gives investors exposure to the bottleneck before electricity ever reaches the server rack.
| Company | Forward P/E | Revenue Growth | Backlog / Visibility | Primary AI Exposure |
| Eaton | 46.0x | 21.4% | 307 GW / 15 years | Electrical infrastructure, power management, cooling |
| Vertiv | 61.8x+ | 24.1% | $15 billion | Critical power, liquid cooling |
| Quanta Services | 77.6x | 41.1% | $53.4 billion | Grid, transmission, substations |
Key Takeaway
In short, investors don’t need to pick the winning AI model to profit from AI infrastructure. Eaton supplies the electrical backbone, Vertiv handles critical power and cooling inside the facility, and Quanta helps connect enormous new loads to the grid.
Granted, valuations have expanded across this group, and not every announced data center will get built on schedule. But that is precisely why Eaton’s 307-GW, 15-year backlog matters. AI may eventually experience a GPU shortage or a spending slowdown, but the physical infrastructure required to support the computing already planned stretches years into the future.
For investors seeking a broader way to participate in AI than simply chasing the hottest chip stock, these three picks-and-shovels companies deserve a place on the watch list.
Contact [email protected] for any questions or corrections.