The data-center boom is spreading beyond chips
Bloomberg Intelligence’s revised estimates point to a striking shift: its forecast for global data-center capacity in 2027 rose about 10%, from 108 gigawatts to 119, while the projected cost of building that capacity climbed roughly 70%, from about $1 trillion to $1.7 trillion. The gap between those increases is the story. Planned capacity is growing, but the estimated bill is rising much faster.
The facilities aren’t simply getting bigger. Each gigawatt now requires more computing hardware, memory, storage, networking, electricity and cooling. That broader bill reflects how a data center works as a connected system: adding computing capacity can also increase demand for the equipment and power needed to run it. A larger capacity forecast, by itself, therefore doesn’t capture the scale of the investment challenge.
That makes this more than a bigger order book for chipmakers. The spending reaches across the equipment inside a data center and the power infrastructure outside it. Chips may draw much of the attention, but servers need memory and storage, networks must move data, and cooling systems must manage the heat generated by intensive computing. The buildout draws on several linked supplier categories.

Why the cost estimate is rising faster than capacity
The estimates suggest that a gigawatt of planned capacity now carries a higher associated construction cost than the earlier forecast implied. The source points to the reason: each gigawatt requires more of the equipment and infrastructure that supports computing. That can make capacity growth more expensive even when the increase in the capacity forecast appears comparatively modest.
Here’s the thing: the estimate describes a projected market, not guaranteed spending or a fixed cost for every project. Actual investment depends on whether planned facilities are built, what equipment they require, and whether customers are prepared to pay for the capacity. Similar infrastructure projects typically involve multiple interdependent costs, so pressure in one part of the system can affect the economics of the whole buildout.
This is more significant than it looks. A rising bill may create opportunities for suppliers, but it also raises the amount operators must commit before a facility can generate revenue. If you’re working in this space, the distinction between demand for equipment and profitable demand for completed capacity matters.
Suppliers and operators face different risks
Bloomberg’s model puts the market near $1.7 trillion in 2027. The central question is who can turn that spending into returns. Data-center operators must fund construction, find customers for the capacity and earn a return over time. Suppliers can be paid as projects are built, which gives them a different risk profile. That distinction matters as the bill rises.
A supplier’s opportunity is tied to equipment orders, but an order book alone doesn’t show whether a project will produce attractive returns for the operator. Operators carry a longer chain of obligations: they need financing for construction, enough customer demand to use the new capacity, and revenue over time to support the investment. Those risks can persist even when the broader market forecast points upward.
And yet, the supplier position isn’t automatically secure. Demand for equipment depends on projects proceeding, while the operator’s ability to fund and fill facilities shapes the pace of construction. The expected spending could be spread across chips, memory, storage, networking, cooling and power infrastructure, but the source doesn’t identify which suppliers will capture the largest share or how their returns will compare.
What the forecast leaves open
The revised estimate is a useful measure of the scale of the planned buildout, but it doesn’t settle whether that investment will earn a sufficient return. The figures are forecasts for 2027, and they describe global capacity and estimated construction costs. They don’t establish which projects will be completed, how quickly customers will take up capacity, or how the spending will be divided among suppliers and operators.
(And this is the part most people overlook.) A market can grow while individual participants face very different outcomes. Suppliers may receive revenue during construction, while operators need to carry the assets and secure customers over a longer period. That difference helps explain why a large projected market isn’t, by itself, proof that every company connected to it will benefit.
The numbers here are underwhelming if read as a simple growth story: capacity rises about 10%, while projected building costs climb roughly 70%. Read another way, they point to a broader and more demanding infrastructure cycle. More than chips. The outlook will depend on whether the additional spending supports useful capacity and whether operators can turn that capacity into sustained returns.
About the analyst
Luke Lango is the editor of Hypergrowth Investing. His work focuses on finding early investment opportunities in fast-growing industries and major long-term trends. You can learn more about Luke.
This article was printed from InvestorPlace Media: “The AI Data Center Bill Keeps Rising: These 7 Stocks Could Collect”. ©2026 InvestorPlace Media, LLC.