Global spending on data centers is forecast to reach $2.1 trillion by 2028, more than twice the $1 trillion expected in 2026, according to new research from Bank of America Global Research. The hyper-speed expansion, driven by demand for AI computing, will create numerous opportunities for chipmakers, server vendors and industrial gear manufacturers.

The bank’s October 23 report predicts that the global data center market, valued at a mere $275 billion in 2022, could reach $2.93 trillion by 2030. AI demand is expected to fuel a torrid compound annual growth rate (CAGR) of 37% between 2025 and 2030.

The forecast highlights the gargantuan investment needed to support AI model training and inference, both of which rely on next-gen processors and leading-edge computing infrastructure.

The growth in data center spending extends well beyond the GPUs that drive frontier model performance. Electrical equipment, networking hardware and cooling systems are also essential components of these facilities, creating a much larger market for infrastructure vendors.

Servers Dominate Data Center Costs

Bank of America estimates that developing data center capacity costs approximately $49 million per megawatt, with IT equipment accounting for 78% of total spending.

Leading chipmakers like NVIDIA and AMD are expected to benefit greatly from the buildout. Server manufacturers are also positioned to capture major spending as tech companies expand their computing capacity, particularly in the U.S.

Indeed, servers represent by far the largest expense, at approximately $31.4 million per megawatt. Networking equipment adds another $4.8 million per megawatt.

These figures demonstrate the core role of computing hardware in data center economics. For companies building AI infrastructure, purchasing enough processing capacity is a much larger expense than many of the physical systems required to house and run that equipment.

The economics are even more demanding for specialized AI infrastructure. Bank of America estimates that neocloud facilities, which focus on providing computing resources for AI workloads, cost approximately $57 million per megawatt. These facilities require more expensive infrastructure than conventional data centers, although their designs can reduce the amount of physical space needed.

Opportunities for Supporting Players

While servers account for most data center costs, the expansion also creates substantial demand for traditional industrial equipment. These systems support the reliable operation of AI computing equipment, which requires enormous electricity and produces significant heat. Bank of America estimates that individual infrastructure components, including generators, electrical distribution equipment, switchgear and cooling systems, cost between $100,000 and $1.1 million per megawatt.

For industrial manufacturers, the data center market offers a new source of revenue tied directly to tech investment. Construction companies and engineering firms also stand to benefit from the buildout of additional facilities.

Data center developers must secure consistent electricity supplies, connect facilities to power grids and obtain construction approval, all of which present growing challenges. Adding to the difficulties, local resistance to data center construction is creating big hurdles for developers.

The question of investment returns seems to be growing more urgent. Tech companies are committing huge sums to infrastructure, but the long-term economics depend on generating sufficient revenue from AI services. When and if this will fully materialize remain open questions.