The global artificial intelligence infrastructure boom is entering a potentially decisive phase, with data center construction expected to accelerate sharply in 2027 as technology companies compete for computing capacity. However, the projected expansion also raises questions about electricity availability, construction timelines and whether future AI revenues will justify the scale of capital investment.

Data Center Capacity Could More Than Double

Annual additions to data center IT capacity outside China are projected to increase from approximately 25 gigawatts in 2026 to 51 gigawatts in 2027. If achieved, the increase would represent a significant acceleration in the physical infrastructure supporting generative AI, cloud computing and increasingly sophisticated digital services. The projected additions for 2027 could approach the scale of the entire global data center build-out completed through 2024, illustrating how rapidly the industry is expanding.

The forecast nevertheless remains conditional. Approximately three-quarters of the anticipated capacity is associated with facilities already operational or under construction, while the remainder consists of projects still in the planning phase. More than 24 gigawatts across 40 campuses are reportedly in advanced construction for potential delivery next year. Even with this progress, permitting, equipment availability and power connections could determine how much capacity actually comes online.

Modular Construction Creates New Opportunities

To accelerate delivery, developers are increasingly adopting modular construction methods. Electrical rooms, cooling systems and other essential components can be manufactured off-site while construction crews prepare the main facility. This approach may reduce the time required for complex mechanical and electrical installation, which can otherwise delay the launch of high-performance computing infrastructure.

A modeled 50-megawatt data center project illustrates the potential benefit: conventional construction could take approximately 22 months, compared with around 14 months using a fully modular approach, excluding permitting. Earlier deployment could allow operators to begin generating revenue sooner and help customers secure scarce computing capacity.

The implications extend beyond semiconductor manufacturers. Suppliers of power-management systems, cooling equipment, prefabricated infrastructure and engineering services could benefit from sustained investment. However, the commercial gains will depend on project execution, pricing pressure and the ability of suppliers to meet demand without excessive cost overruns.

Power Constraints and Returns Remain the Main Risks

Accelerated construction does not guarantee operational readiness. Data centers require buildings, specialized chips, reliable electricity and grid connections to become productive assets. Delays in any one of these areas can leave substantial capital tied up without generating expected returns. Recent infrastructure decisions have underscored how power-delivery timelines can alter even projects backed by major technology companies.

For investors, the central question is increasingly whether infrastructure spending will translate into durable AI-related revenue and stronger cash flow. Higher borrowing costs, financing commitments and competition for energy resources could weaken project economics, particularly for operators relying heavily on debt. Regulatory hurdles, community opposition and slower-than-expected AI adoption present additional downside risks.

Outlook: Execution Will Matter More Than Announced Capacity

During the remainder of the week and into 2027, investors will be watching construction milestones, power agreements, capital expenditure guidance and evidence that AI demand is supporting profitable services. The projected expansion could create opportunities across the technology and industrial supply chains, but the 51-gigawatt forecast should be treated as a scenario rather than a guaranteed outcome. For global asset allocators, including Israeli institutional investors exposed to international technology markets, the distinction between announced capacity and operational, revenue-generating infrastructure will remain critical.