Artificial intelligence is turning the world's data centers into massive construction sites, and the bill is climbing faster than many expected. Bank of America Global Research now projects global data center spending will hit $1 trillion in 2026 and $2.1 trillion by 2028, with servers taking the largest slice of that spending.
The bank's analysts, in a recent note, argue that AI is fundamentally changing what a data center is. Instead of being primarily real estate projects—buildings that house servers—they are becoming compute-heavy infrastructure builds where the technology inside matters more than the shell around it.
From $275 billion to nearly $3 trillion
To put the scale in perspective, Bank of America estimates data center spending was $275 billion in 2022. By 2030, it could reach $2.93 trillion, driven by rapid growth later this decade. That would represent more than a tenfold increase in less than a decade.
The cost mix is telling. The bank pegs total build costs at roughly $49 million per megawatt of capacity. Of that, information technology equipment—primarily servers, networking gear, and storage—makes up about 78% of the total. That means the traditional construction costs, like land, buildings, and cooling systems, are now a minority share.
This shift explains why companies like ABB, which supplies electrical equipment for data centers, have seen record backlogs. The demand for power infrastructure, cooling, and specialized components is surging alongside the AI boom.
Why servers dominate the bill
Servers are the workhorses of AI. Training large language models and running inference—the process of generating responses—requires enormous computing power. That means data centers need thousands of high-end graphics processing units (GPUs) and custom chips, which are far more expensive than traditional server CPUs.
As AI models grow larger, the need for compute scales up. This is why the spending mix has tilted so heavily toward IT equipment. A data center's value is increasingly measured by the number of chips it can house and cool, not just its square footage.
This trend has ripple effects across the tech supply chain. Chipmakers, server manufacturers, and networking companies are all seeing demand surge. At the same time, some software firms are feeling pressure as AI shifts spending priorities—Infosys recently trimmed its growth outlook, citing AI pricing pressure in North America, a sign that not every tech company benefits equally.
What it means for investors
For everyday investors, the key takeaway is that AI infrastructure spending is not a niche story—it is becoming a macroeconomic force. The scale of investment is comparable to major historical buildouts, like the expansion of the interstate highway system or the fiber-optic boom of the early 2000s.
But that comparison also carries a warning. The fiber-optic boom eventually led to overcapacity and a bust. While AI demand is currently robust, investors should be aware that such rapid spending cycles can be volatile.
Companies that supply the building blocks—semiconductors, cooling systems, power management, and construction services—are likely to see strong revenue growth in the coming years. However, valuations for many of these names are already elevated, reflecting high expectations.
For those who own broad index funds, this spending is already a major driver of earnings growth for tech-heavy indices. For individual stock pickers, the challenge is distinguishing between companies with durable competitive advantages and those that are simply riding the wave.
As the buildout progresses, watch for signs of demand softening. If AI adoption slows or if efficiency gains reduce the need for new data centers, the trillion-dollar projections could be revised down. Conversely, if AI applications continue to expand, the numbers could prove conservative.
For now, the trajectory is clear: AI is not just a software story—it is a hardware and infrastructure story, and the spending is only getting bigger.
