Processing power is moving away from centralized data centers faster than most forecasts predicted — but the numbers show two things happening at once, not one.

The shift toward edge and distributed computing is being driven less by any single breakthrough than by an accumulation of smaller pressures: latency-sensitive applications, rising energy costs, and regulatory requirements that keep data within national borders. At the same time, the biggest cloud providers are pouring unprecedented sums into ever-larger centralized facilities to train and run artificial intelligence models. Both trends are real, and both show up clearly in the market data.

Edge Spending Is Growing Faster Than the Cloud Market Overall

According to IDC’s Worldwide Edge Spending Guide, global spending on edge computing — hardware, software, and services that process data closer to where it is generated rather than in a distant data center — reached nearly $261 billion in 2025. IDC projects that figure will grow at a compound annual growth rate of 13.8 percent, reaching close to $380 billion by 2028. Within that spending, IDC expects provisioned services to grow faster than hardware purchases, with services posting a five-year compound annual growth rate above 18 percent — a sign that companies increasingly buy edge computing as a managed capability rather than assembling it themselves.

Separately, telecom-focused market analysts have tracked roughly $100 billion in expected service-provider investment by 2028 in multi-access edge computing, content delivery networks, and virtual network functions — the infrastructure that lets carriers push computing power toward cell towers and regional hubs instead of routing every request back to a distant cloud region. Retailers, factories, and telecom operators have been among the fastest adopters, largely because point-of-sale systems, factory-floor sensors, and 5G networks generate data faster than it can usefully travel round-trip to a centralized cloud.

Yet the Centralized Side of the Market Is Also Exploding

The counterintuitive part of the story is that centralized infrastructure is not shrinking — it is growing at a pace edge computing has not matched. Synergy Research Group’s tracking of hyperscale data centers found 1,136 such facilities operating worldwide by the end of 2024, with 137 added that year alone and another 130 to 140 expected annually going forward. Synergy counted 504 additional hyperscale facilities already planned, under construction, or being fitted out. The firm noted that the world’s hyperscale data center capacity — measured in megawatts of critical IT load — has been doubling roughly every four years, even as the number of facilities merely doubled over five years, meaning each new generation of data center is markedly larger than the last.

“The big difference now is the increased scale of many of those new data centers,” John Dinsdale, Synergy’s chief analyst, said of the trend, attributing it largely to the buildout of AI infrastructure. Amazon, Microsoft, and Google together account for 59 percent of all hyperscale data center capacity worldwide, and the United States alone holds 54 percent of global capacity, with Europe and China splitting most of the remainder.

What Is Actually Driving the Power Curve

The reason both trends can be true simultaneously shows up in electricity data. The International Energy Agency estimated that data centers consumed around 415 terawatt-hours of electricity in 2024 — about 1.5 percent of global electricity use — and projected that figure would climb to roughly 945 terawatt-hours by 2030, or nearly 3 percent of global consumption. The IEA attributes most of that acceleration to AI: it expects electricity demand from AI-optimized servers to grow around 30 percent annually through the decade, versus roughly 9 percent for conventional servers, with the United States adding about 240 terawatt-hours of new data center demand (a 130 percent increase) and China adding about 175 terawatt-hours (a 170 percent increase) between 2024 and 2030.

Put together, the data describes two distinct migrations rather than one. Latency-sensitive, everyday computing — retail transactions, industrial sensors, video streaming, connected vehicles — is genuinely moving outward, toward the edge, and IDC’s numbers show that shift accelerating. But the heaviest, most power-hungry workloads, above all AI model training and inference, are consolidating into fewer, far larger hyperscale campuses than existed five years ago. The map of where computing happens is not simply flattening or simply centralizing. It is doing both, for different kinds of work, at the same time.

Sources: IDC, “IDC Estimates Global Spending on Edge Computing to Grow at 13.8% Reaching Nearly $380 Billion by 2028,” press release (idc.com); Synergy Research Group, “Hyperscale Data Center Count Hits 1,136; Average Size Increases; US Accounts for 54% of Total Capacity” (srgresearch.com); International Energy Agency, “Energy Demand from AI,” Energy and AI report (iea.org).