What is a hyperscaler?

Is there an official definition of a hyperscaler?

There is no single official definition of the word. The most cited count comes from Synergy Research Group, an industry analyst firm, which tracks 19 large cloud and internet companies it treats as hyperscale operators[1], spanning cloud services, search, social networking, e-commerce and gaming[2]. Synergy does not list its full criteria in its releases.

In practice the word is used two ways. Narrowly, it means the few companies that run their own global networks of very large data centers. Loosely, it covers anyone building at gigawatt scale, including AI labs and the developers that build for them. This guide uses the narrow sense. For a size-based definition, see the glossary entry on hyperscale data centers.

How many hyperscale data centers are there?

Synergy counted 1,136 hyperscale data centers worldwide at the end of 2024[1], run by those 19 companies[1]. It said the United States holds well over half of their capacity[1]; its headline puts the share at 54%, measured by critical IT load, the power available to computing equipment.

Synergy’s count filters out smaller points of presence[1], so it is a count of large sites, not of every facility these companies use.

What is the difference between a hyperscaler and a hyperscale data center?

A hyperscaler is the company. A hyperscale data center is a building or campus large enough to serve one. Synergy counts only large data centers that hyperscale operators run[1]. Developers also use the word for large campuses they build to lease to these companies, so “hyperscale” on a rezoning application does not by itself say who the tenant is.

Who are the hyperscalers, and how much do they spend?

Amazon, Microsoft, Alphabet, Meta and Oracle reported the capital expenditures below in their latest SEC filings. Each ↗ links to the filing.

Total capital expenditures as each company reports them, including spending on property and equipment beyond data centers. Fiscal calendars differ, so the rows are not like-for-like.

For quarter-by-quarter history, see the AI capex tracker. CoreWeave, a neocloud, is tracked there too.

Where are hyperscalers building in the U.S.?

The table counts the Census’s U.S. project records that name one of the five as developer, operator or end user (developer or end user).

Amazon has the most tracked projects (19); Meta has the most planned capacity (10.42 GW, active projects).

Many hyperscaler campuses are built by developers and leased, and the end user is not always public. Projects whose end user has not been disclosed are not counted here.

Hyperscaler, neocloud or colocation provider: what’s the difference?

More on each: What is a neocloud? and What is a colocation data center?

Why does it matter which hyperscaler is behind a project?

Because the end user shapes what gets built and what a community is asked to approve: how much power, how fast, and what kind of cooling. Many proposals reach a planning board under a code name or a newly formed company before the end user is named. The Census’s code-name list matches 52 code names to the companies behind them. For the overall U.S. picture, see How many data centers are in the U.S.? and the Stargate tracker. Oracle, for example, agreed with OpenAI in July 2025 to develop 4.5 gigawatts of additional Stargate capacity, which is why it is often grouped with the hyperscalers.

Common questions

Is Oracle a hyperscaler?

Oracle runs Oracle Cloud Infrastructure and is often grouped with the hyperscalers, more so since it agreed with OpenAI in July 2025 to develop 4.5 gigawatts of additional Stargate capacity.

Is OpenAI a hyperscaler?

Not in the usual sense. OpenAI relies on partners for its data centers, including Oracle, which is developing Stargate sites with it, and SB Energy, which builds and operates its Milam County, Texas, site. See the Stargate tracker for each site.

Is Nvidia a hyperscaler?

No. Nvidia designs the GPUs that hyperscalers buy; it is counted as a supplier, not a hyperscale operator. GPU rental prices on the Census show what those chips cost to rent.

How many hyperscalers are there?

It depends on the definition. Synergy Research tracks 19 companies as hyperscale operators. In everyday use, the word usually means five: Amazon, Microsoft, Google, Meta and Oracle.

What is hyperscale computing?

Computing built to grow by adding large numbers of standard servers, storage and network gear, the way the biggest cloud companies run their data centers.

Sources

1. Hyperscale Data Center Count Hits 1,136; Average Size Increases; US Accounts for 54% of Total Capacity
   Synergy Research Group · 2025-03-19
   4 source excerpts

   “19 of the world's major cloud and internet service firms”

   “increased to 1,136 at the end of 2024”

   “well over half of total worldwide capacity”

   “filtered out smaller points of presence”

2. Hyperscale Data Center Capacity to Almost Triple in Next Six Years, Driven by AI
   Synergy Research Group · 2023-10-17
   Source excerpt

   “SaaS, IaaS, PaaS, search, social networking, e-commerce and gaming”

Each source was opened and checked against the excerpt shown. Company statements are identified as such. Figures are as published; where sources differ, each is shown.