Amazon Web Services retained the lead in global data center capacity in 2026, with Microsoft, Google, Meta, Oracle, and Alibaba forming the principal hyperscale expansion group [S3][S7].
The largest operators—AWS, Microsoft, Google, Meta, Oracle, and Alibaba—were expanding aggressively, with their capital expenditures projected to exceed USD 600 billion in 2026, while hyperscale sites reached 1,360 and represented 48% of worldwide data center capacity at the end of 2025 [S1][S3].
Capacity leadership is measurable, but not a single global ranking
AWS remained the global capacity leader in 2026 [S7]. Synergy tracked 1,360 operational hyperscale data centers at the end of the fourth quarter of 2025 and measured their share of worldwide data center capacity at 48% [S1]. Its earlier count was 1,297 operational hyperscale data centers as of late 2025 [S3].
The practical ranking cannot be reduced to facility count. Synergy estimates that almost 60% of hyperscale capacity sits in operator-built, operator-owned facilities, with the remaining 40% leased [S1]. Capacity therefore depends on whether the metric counts campuses, buildings, halls, IT load, or leased footprint.
A hyperscale data center can occupy as much as 10 million square feet, compared with an average data center footprint of about 100,000 square feet [S2]. The same estimate set places 13,030 active data centers across 181 countries as of October 2026, including 5,044 in the United States, 595 in the United Kingdom, 580 in Germany, 446 in China, and 401 in France [S2].
AWS leads on footprint, Microsoft on cloud regions
The operator footprints show different expansion models.
These figures explain why a facility-count ranking alone can mislead buyers evaluating redundancy, cooling architecture, network reach, and contracted capacity.
Meta and Alibaba sit within the same leading tier
AWS, Microsoft, Google, Meta, Oracle, and Alibaba were identified as the largest hyperscale operators [S3]. Their inclusion matters because cloud market share, capital spending, and data center capacity are not interchangeable measures. A provider may operate a large cloud business while using leased capacity, or own large campuses that serve third-party or internal workloads.
The hyperscale market was valued at USD 318.70 billion in 2025 and estimated at USD 473.76 billion in 2026, with a forecast increase to USD 1,869.27 billion by 2035 [S5]. Servers and compute represented about 40.5% of the hyperscale data center market in 2025 [S5].
Colocation providers rank differently from hyperscale owners
They are major data center operators, but their business model differs from AWS, Microsoft, Google, Meta, Oracle, and Alibaba because they lease capacity rather than primarily consuming it for cloud and platform services.
These counts place the firms on infrastructure-provider rankings, not the hyperscale cloud-owner ranking.
For capacity decisions, the comparison is straightforward: hyperscale owners control cloud-platform demand but may depend on leased sites; colocation operators own or lease facilities and serve multiple customers; enterprise on-premises capacity offers direct control but carries the customer’s own build and operating burden. Hyperscale capacity represented 48% of the global market at the end of 2025, non-hyperscale colocation represented 20%, and enterprise on-premises facilities represented 32% [S1].
Power and delivery risk can override headline rankings
Power access had overtaken land and capital as the binding constraint on new hyperscale capacity, and construction schedules rather than order books determine the pace of supplier delivery [S5]. As of June 2025, more than 36 projects representing USD 162 billion in investment had been blocked or significantly delayed [S3].
Cooling and electrical demands are also changing the supply chain. Data center investment through 2030 was projected at nearly USD 7 trillion, with about 15% allocated to land and buildings, 25% to power generation and cooling, and 60% to technology and hardware [S2]. The supply chain for DC power supplies, switchgear, cooling, and other capacity-critical equipment therefore matters as much as the operator ranking.
Project delays separate announced pipelines from delivered capacity
Synergy counted almost 800 hyperscale facilities in its known future pipeline, allowing capacity to double in three years, while the January 2026 market review listed 770 facilities at planning, construction, or fit-out stages [S1][S3]. The difference between pipeline and operational capacity is the key qualification behind any 2026 ranking.
Hyperscale operators were expected to hold 67% of global data center capacity by 2031, while on-premises share was projected to fall to 19%, down from 56% in 2018 [S1]. Colocation capacity was expected to keep growing at close to double-digit annual rates even as its share declined [S1]. The structural allocation is moving toward hyperscale and colocation, but announced megawatts do not become dependable capacity until power, cooling, fiber, commissioning, and customer acceptance are complete.
Additional context for equipment planning is available in Hyperscale Colocation Leasing: 2026 Capacity and Contract Scale.
The trackable signals for operator ranking are delivered IT capacity, energized megawatts, operational facility count, and signed leases rather than announced campus value. By those measures, AWS remained the capacity leader in 2026, with Microsoft, Google, Meta, Oracle, and Alibaba forming the leading hyperscale group, while 1,360 operational sites and 48% market share established the sector baseline entering 2026 [S1][S3][S7].