Bank of America Forecasts Global Data Center Spending to Reach $2.1 Trillion by 2028
nashnova research
Bank of America projects global data center spending will surpass $2.1 trillion by 2028, driven by AI deployment; this means a market worth just $275 billion in 2022 could expand nearly eightfold in six years.
How fast is this market really growing?
BofA expects the data center market to break $1 trillion in 2026, exceed $2.1 trillion in 2028, and reach $2.93 trillion by 2030.
That implies a compound annual growth rate of roughly 37% from 2025 to 2030.
This means → data centers are shifting from "IT infrastructure" to a super-cycle growing at early-internet speed — and AI is the only variable that explains the pace.
Where does the money go when you build a data center?
BofA estimates construction costs at roughly $49 million per megawatt, with IT-related spending accounting for 78% of the total.
Servers are the single largest line item at about $31.4 million per MW; networking equipment ranks second at roughly $4.8 million per MW.
In plain terms = nearly eight out of every ten dollars go to "compute itself" — chips and servers. Power, cooling, generators, and backup combined take the remaining two.
What makes an AI-specific data center different?
BofA calls AI-focused facilities the "new cloud," estimating their build cost at about $57 million per MW — roughly 16% more than a traditional data center.
However, AI data centers need less physical space, partly offsetting the higher per-unit cost.
This reflects a density trade-off: more chips packed into the same footprint raises the cost per megawatt but shrinks total acreage.
Who profits from this spending wave?
The report names chipmakers and server developers as the most direct beneficiaries, with gains concentrated in the U.S. market.
Makers of generators, switchgear, and cooling systems — industrial equipment manufacturers — are also booking data centers as a meaningful revenue source.
This means → the beneficiary chain extends from "core compute" outward to "power + thermal." The closer a company sits to the server, the higher its certainty; the closer to civil construction, the greater the optionality.
How reliable is this forecast?
BofA itself flags that whether AI capital expenditure materializes on schedule before 2028 is the critical checkpoint for the entire supply-chain thesis.
In plain terms = $2.1 trillion is not money already spent — it is a projection built on the assumption that AI deployment keeps scaling. If AI adoption disappoints, every link in this chain gets marked down.
The current certainty ranking: servers > networking > power and cooling > civil works. The closer to the chip, the less likely the budget gets cut.
市场有风险，内容仅供研究参考，不构成投资建议。