15 Largest Data Centers in the World

Every app you open, video you stream, or file you save is backed by a data center somewhere. But some aren’t just big, they are absolute monsters.

These facilities are the backbone of modern tech, built to handle everything from cloud storage to national security. Let’s cut through the noise and look at the biggest data centers in the world and what makes them impossible to ignore.

This list ranks the biggest data centers in operation today by total reported floor area. By that measure, the largest is China Telecom’s Inner Mongolia Information Park in Hohhot, at roughly 10.7 million square feet. 

The list covers existing, operating facilities only, not projects still under construction. Reported floor areas for several of the largest campuses come from operators or state media.

What Is the Biggest Data Center in the World?

The China Telecom Inner Mongolia Information Park near Hohhot, China, is the world’s biggest data center.

The complex covers approximately 1 million square meters, or 10.7 million square feet, and has a reported power capacity of around 150 megawatts. The total area includes data halls and supporting buildings, so the exact amount of active server space is not publicly confirmed.

By projected computing capacity, Meta’s Hyperion data center campus in Louisiana may eventually take the lead. Meta expanded the project to a planned 5 gigawatts, but the campus remains under construction and cannot yet be classified as the largest operational data center.

OpenAI’s Stargate campus in Abilene, Texas, is the flagship of a $500 billion program backed by OpenAI, Oracle, and SoftBank. Developed by Crusoe, the site covers about 4 million square feet and is designed for roughly 1.2 gigawatts. About half the campus was operational by mid-2026, while the remaining buildings were still under construction.

15 Largest Data Centers in the World

Here’s a detailed list of the 15 largest data centers worldwide, ranked from the smallest to the largest. The biggest data center is at the bottom, so keep scrolling to see it!

For a quick comparison, we’ve also included a table below:

15 Largest Data Centers in the World Comparison Table

Rank	Data Center / Campus	Location	Reported Area	Power
15	NAP of the Americas (Equinix MI1)	USA	750,000 sq ft	17 MW
14	Yotta NM1	India	820,000 sq ft	50 MW
13	QTS Metro Data Center	USA	970,000 sq ft	70 to 130 MW
12	Tulip Data Center	India	1.0M sq ft	100 MW
11	Lakeside Technology Center	USA	1.1M sq ft	~100 MW
10	Apple Mesa Data Center	USA	1.3M sq ft	~50 MW
9	Utah Data Center	USA	1.4M sq ft	65 MW
8	ACC Campus	USA	1.6M sq ft	42+ MW
7	The Citadel Campus	USA	~1.8M sq ft operational (7.2M master plan)	650 MW planned
6	CWL1 Data Centre	UK	2.0M sq ft	~148 MW
5	Switch Core Campus	USA	2.7M sq ft	~315 MW
4	Google Council Bluffs Data Center	USA	2.9M sq ft	100+ MW
3	Harbin Data Center	China	7.1M sq ft	~200 MW
2	China Mobile Hohhot Data Center	China	7.7M sq ft	150+ MW
1	China Telecom Inner Mongolia Information Park	China	10.7M sq ft	~150 MW

15. NAP of the Americas (Equinix MI1)

• Company: Equinix
• Location: Miami, Florida, USA
• Power Capacity: 17 MW
• Area: 750,000 sq. ft
• Cost: Undisclosed
• Operator: Equinix
• Cabinet Capacity: 6,489 cabinets; configs from 5 kVA
• Cooling System: N+1 redundancy; free cooling; hot/cold aisle containment
• Renewable Energy Use: US/Canada wind and solar RECs; Texas/Oklahoma vPPAs

Now operated as Equinix MI1, the NAP of the Americas is a 750,000-square-foot, six-story data center and one of the main internet gateways connecting the United States with Latin America and the Caribbean. Located in downtown Miami, the carrier-neutral facility supports telecom providers, cloud platforms, financial institutions, government agencies, content companies, and international enterprises.

Equinix lists 255,513 square feet of active colocation space and capacity for 6,489 cabinets, with cabinet configurations starting at 5 kVA. The facility uses N+1 power and cooling redundancy, provides more than 30 hours of generator autonomy at full load, and supports floor loads of 250 pounds per square foot.

MI1 connects customers to more than 130 carriers, regional submarine cable systems, cloud providers, Equinix Fabric, Internet Exchange services, and metro fiber networks. Reinforced concrete construction, biometric access controls, security mantraps, continuous monitoring, and placement above the 500-year flood elevation protect the facility from physical and environmental risks.

14. Yotta NM1 (Navi Mumbai, India)

• Company: Yotta Infrastructure
• Location: Panvel, Navi Mumbai, India
• Power Capacity: 50 MW (expandable to 60 MW)
• Area: 76,180 sq. meters (820,000 sq. ft)
• Cost: $1.1B+ (entire park)
• CEO: Sunil Gupta
• Rack Capacity: Up to 7,200 racks
• PUE: Design PUE of 1.4
• Cooling System: Chilled-water cooling with N+1 or N+2 chiller redundancy
• Tier Rating: Tier IV TCDD, TCCF, and Gold Operational Sustainability
• Renewable Energy Use: Captive solar and renewable sources planned
• Rack Space: Up to 7,200 racks

Yotta NM1 is a 16-story hyperscale data center covering 820,000 square feet in Navi Mumbai. Rated at 50 MW and expandable to 60 MW, its operational profile lists 52 MW of IT capacity, support for up to 7,200 racks, redundant power and cooling, fiber connectivity, and 48 hours of backup power.

NM1 was India’s first data center to receive Uptime Institute Tier IV certifications for both design and constructed facility standards. It later earned Tier IV Gold operational sustainability certification. The carrier-neutral site supports banking, telecom, government, cloud, AI, and high-performance computing workloads.

NM1 is the first building in Yotta’s Navi Mumbai data center campus. The original five-building plan targeted 30,000 racks and 250 MW, with additional expansion planned across the site.

13. QTS Metro Data Center (Atlanta, Georgia, USA)

• Company: QTS Realty Trust (owned by Blackstone)
• Location: Atlanta, Georgia, USA
• Power Capacity: 70 to 130 MW
• Area: 90,116 sq. meters (970,000 sq. ft)
• CEO: Chad Williams (Founder)
• Cooling System: Air-side economizers; rainwater collection support; 950 CRAH units
• Renewable Energy Use: ~350 MW solar via Georgia Power agreement
• Rack Space: 527,000 sq. ft of raised-floor space

Converted from a former Sears-Roebuck distribution building, the QTS Atlanta-Metro facility covers 970,000 square feet. The site is commonly listed at 70 to 130 MW, while its original specifications identify 527,000 square feet of raised floor space, 120 MW of critical power expandable to 160 MW, and N+1 to 2N redundancy. Cooling infrastructure includes 10 chillers, 16 cooling towers, 950 CRAH units, air-side economizers, and generators supported by 200,000 gallons of on-site fuel.

The facility is part of QTS Atlanta 1, a 99-acre, four-building campus with more than 278 MW of planned critical capacity and three expandable Georgia Power substations. Redundant fiber routes connect the buildings to multiple carriers, internet exchanges, major cloud platforms, and the 56 Marietta carrier hotel.

Phase II holds LEED Gold certification and uses rainwater collection and energy-efficient air-side cooling. Security includes perimeter fencing, gated access, continuous staffing, video surveillance, and biometric controls. The site serves hyperscale companies, enterprises, government agencies, financial institutions, and healthcare providers.

Blackstone acquired QTS for $10 billion in 2021. Founder Chad Williams previously served as CEO, while Thomas A. “Tag” Greason and David Robey now lead the company as co-CEOs.

12. Tulip Data Center (Bengaluru, India)

• Company: Tulip (data center services now associated with STT GDC)
• Location: Bengaluru, Karnataka, India
• Power Capacity: 100 MW (design)
• Area: 92,900 sq. meters (1 million sq. ft)
• Cost: $600 million
• Operator: Tulip / STT GDC
• Rack Capacity: Up to 12,000 racks
• PUE: 1.5
• Cooling System: 17,000 TR capacity; ambient-air cooling capability
• Tier Rating: Designed to Tier III and IV standards
• Rack Space: Up to 12,000 racks

Opened in Bengaluru as one of India’s largest data centers, the Tulip facility covers roughly 1 million square feet and has a design capacity of 100 megawatts.

The four-tower facility covers 900,000 square feet and includes 20 modular sections designed for up to 12,000 racks and 100 MW of power. IBM advised on its power, cooling, rack, cloud, and operational infrastructure.

The carrier-neutral facility was designed to Tier III and Tier IV standards and launched with a reported PUE of 1.5. It provides colocation, cloud, hosting, storage, security, connectivity, and disaster recovery services.

Engineering records list a 40 MVA electrical demand and 17,000 TR of cooling capacity. Tulip acquired the site for ₹230 crore, with total planned investment reaching ₹900 crore.

11. Lakeside Technology Center (Chicago, Illinois, USA)

• Company: Digital Realty
• Location: Chicago, Illinois, USA
• Power Capacity: 100 MW
• Area: 102,193 sq. meters (1.1 million sq. ft)
• Cost: Estimated cumulative value around $2 billion
• CEO: Andy Power (CEO of Digital Realty)
• Network Scale: 70+ providers; 500+ tenants
• PUE: N/A
• Cooling System: N+1 redundancy; brine-water thermal storage supplies 50% cooling
• Tier Rating: Tier III design; Uptime Institute certification is N/A
• Renewable Energy Use: 100% renewable-energy matching for colocation portfolio
• Rack Space: 206,100 sq. ft built-out space

Housed at 350 East Cermak, the Lakeside Technology Center is one of North America’s largest carrier hotels, with 1.1 million square feet of floor space across eight stories. The facility has 109 MW of utility capacity, 70 MW of UPS capacity, 150 MW of generator capacity, and redundant power configurations ranging from N+1 to 2N.

Built in 1912 as an R.R. Donnelley printing plant, Digital Realty acquired the property in 2005 and converted it into a major internet, cloud, content, and financial trading hub. The facility connects to more than 70 network providers and serves over 500 tenants, including companies requiring low-latency access to Chicago’s financial district.

A refrigerated brine-water system supplies 50% of the building’s cooling requirements. The site holds LEED Gold, Energy Star, SOC 2, SOC 3, PCI DSS, and ISO 27001 credentials. The $2 billion figure represents the property’s investment valuation as of August 2022 rather than its construction cost.

The building delivers approximately 100 megawatts and houses financial institutions, healthcare providers, content platforms, and data center operators, including Equinix and ColoHouse.

10. Apple Mesa Data Center (Mesa, Arizona, USA)

• Company: Apple
• Location: Mesa, Arizona, USA
• Power Capacity: 50 MW (renewable)
• Area: 120,773 sq. meters (1.3 million sq. ft)
• Total Investment: Over $2 billion
• CEO: Tim Cook
• Site Area: 83 acres
• Cooling System: Water-based cooling with fan-assisted server halls
• Renewable Energy Use: 100% renewable; supported by 50 MW Bonnybrooke solar

Apple’s Mesa data center covers 1.3 million square feet on an 83-acre site. Operating as a data center since 2017, the facility serves as a command hub supporting iMessage, Siri, iCloud, Apple Music, the App Store, and other core services while helping monitor Apple’s wider cloud infrastructure.

The building was originally developed as a solar-panel manufacturing site before being repurposed for sapphire glass production. After that project ended, Apple converted the property into a secure data operation backed by more than $2 billion in long-term investment.

The facility has 50 megawatts of capacity and uses renewable electricity supported by Apple’s Bonnybrooke solar plant near Florence, Arizona. Mesa’s stable grid, fiber connectivity, available land, and low natural-disaster exposure make it well suited for continuous data center operations.

9. Utah Data Center (Bluffdale, Utah, USA)

• Organization: National Security Agency (NSA)
• Location: Bluffdale, Utah, USA
• Power Capacity: 65 MW
• Area: 130,064 sq. meters (1.4 million sq. ft)
• Cost: $1.5 billion
• Primary Data Hall Space: 100,000 sq. ft
• Cooling System: Chilled-water; cooling towers; thermal-energy storage
• Tier Rating: Tier III, concurrently maintainable design
• Rack Space: 100,000 sq. ft raised floor (4 halls)

Completed in 2014 at a cost of $1.5 billion, the Utah Data Center spans 1.4 million square feet at Camp Williams near Bluffdale. Operated by the NSA, the facility stores and processes classified information for U.S. intelligence and cybersecurity operations.

The complex includes 100,000 square feet of primary data halls, supported by administrative buildings, utility infrastructure, cooling systems, storage areas, backup power equipment, and controlled security zones.

The site operates with 65 megawatts of power capacity and includes infrastructure designed for continuous computing, large-scale data retention, and round-the-clock availability. Its Utah location provides reliable grid access, physical separation, and low exposure to major natural disasters.

8. ACC Campus (Ashburn, Virginia, USA)

• Company: Digital Realty (formerly DuPont Fabros)
• Location: Ashburn, Virginia, USA
• Power Capacity: 42+ MW
• Area: 148,600 sq. meters (1.6 million sq. ft)
• Cost: Not disclosed
• CEO: Andy Power (CEO of Digital Realty)
• Technical Scale: 28 rooms; each 1 to 2 MW
• PUE: Calculated annualized design PUE of 1.15
• Cooling System: Evaporative cooling (reclaimed water); 12 chillers; thermal storage
• Renewable Energy Use: Wind and solar available via procurement program
• Rack Space: 28 rooms; ~378 cabinets per room

Originally developed by DuPont Fabros and now operated by Digital Realty, the ACC Campus covers 1.6 million square feet across a multi-building data center complex in Ashburn. The campus serves hyperscale and enterprise tenants across Loudoun County, where the surrounding infrastructure carries more than 70% of global internet traffic.

ACC7 is the largest building on the campus, covering 446,000 square feet with 41.6 megawatts of critical power. It contains 28 computer rooms, each supporting 1 to 2 megawatts, with flexible cabinet layouts and power densities. The facility uses N+2 power redundancy, medium-voltage distribution, and evaporative cooling supported partly by reclaimed water. ACC7 holds LEED Gold certification.

The Ashburn location provides dense fiber connectivity, reliable Dominion Energy service, and Virginia’s sales and use tax exemption for qualifying data center equipment. The campus supports cloud platforms, enterprise systems, and mission-critical workloads across multiple buildings.

7. The Switch Citadel Campus (Tahoe Reno, Nevada, USA)

• Company: Switch
• Location: Tahoe Reno, Nevada, USA
• Power Capacity: 185 MW operational; 650 MW planned
• Area: 1.8M sq. ft operational; 7.2M master plan
• Cost: Not disclosed
• CEO: Rob Roy
• Cabinet Density: Up to 55 kW per cabinet
• PUE: 1.18 to 1.20 annual campus average
• Cooling System: Hot-aisle containment; air-to-liquid exchangers; water-independent capable
• Tier Rating: Switch Tier 5 Platinum, a proprietary Switch standard
• Renewable Energy Use: 100% renewable; net-zero operational carbon

Operated by Switch in the Tahoe Reno Industrial Center, the Citadel Campus has an operational footprint of roughly 1.3 to 1.8 million square feet across two commissioned buildings. The campus sits beside the Tesla Gigafactory and supports enterprise, government, cloud, AI, and high-performance computing workloads.

The lead building, TAHOE RENO 1, spans 1.3 million square feet and provides up to 130 megawatts of capacity. Switch has described it as one of the largest single colocation buildings in operation. The facility supports cabinet densities of up to 55 kW and uses modular data halls, tri-redundant UPS infrastructure, and dual roof decks rated for 200 mph winds. A 20-foot concrete perimeter wall, seven layers of security, military-trained personnel, and an on-site fire brigade protect the site.

TAHOE RENO 2 adds more than 522,000 square feet and 55 megawatts of capacity, with a stated PUE of 1.18. The campus runs on 100% renewable energy, follows Switch’s Tier 5 standard, and carries a 100% power uptime guarantee for supported deployments.

The Switch Superloop delivers fiber connections to San Francisco, the Bay Area, Los Angeles, and Las Vegas, with 4.5-millisecond latency to the Bay Area and a 7-millisecond route to Switch’s Core Campus. DigitalBridge and IFM acquired Switch and took the company private in December 2022. Measured by operational area, the Citadel ranks below Switch’s Las Vegas Core Campus, which operates at 2.7 million square feet.

Switch has master-planned the Citadel for 7.2 million square feet and up to 650 megawatts across more than 2,000 acres, with additional buildings under construction. That figure describes the full build-out rather than the current operational facility.

6. CWL1 Data Centre (Newport, Wales, UK)

• Company: Vantage Data Centers
• Location: Newport, Wales, UK
• Power Capacity: 148 MW; expandable to 180 MW+
• Area: 185,806 sq. meters (2 million sq. ft)
• Cost: Estimated $1.3 billion
• CEO: Sureel Choksi
• Rack Density: 2 kW to more than 125 kW per rack
• Cooling System: Indirect evaporative; liquid-to-chip; free-air cooling
• Tier Rating: Designed and built to exceed Tier III standards
• Renewable Energy Use: 100% renewable electricity
• Rack Space: N/A

CWL1 is one of Europe’s largest data center campuses, spanning 2 million square feet across a 46-acre site near Newport in South Wales. Originally developed by Next Generation Data and acquired by Vantage Data Centers in 2020, the campus is planned for three buildings, with two operational, and provides 148 megawatts of critical IT capacity.

Rack densities range from 2 kW to more than 125 kW, supporting enterprise, cloud, AI, and high-performance computing workloads.

The campus exceeds Tier III standards and uses an N+N power design connected directly to the UK SuperGrid through a private 400 kV substation. Cooling options include free-air cooling, indirect evaporative systems, dry coolers, liquid-to-chip cooling, in-row units, and hot-and-cold aisle containment.

CWL1 runs on 100% renewable electricity and has maintained 100% uptime since 2010. The carrier-neutral site provides latency below 1.5 milliseconds to London, multiple fiber entry points, eight meet-me rooms, Tier 1 carriers, cloud connections, and access to the LINX Wales internet exchange.

5. Switch Core Campus (Las Vegas, Nevada, USA)

• Company: Switch
• Location: Las Vegas, Nevada, USA
• Power Capacity: 315 MW; expandable to 495 MW
• Area: 250,800 sq. meters (2.7 million sq. ft operational)
• Cost: Over $1.1 billion
• CEO: Rob Roy
• Cabinet Density: Up to 55 kW per cabinet
• PUE: Annual PUE of 1.28
• Cooling System: Hot-aisle containment; air-to-liquid exchangers; Tier 5 standards
• Tier Rating: Switch Tier 5 Platinum, a proprietary Switch standard
• Renewable Energy Use: 100% renewable; net-zero operational carbon

The Switch Core Campus in Las Vegas is one of the largest colocation ecosystems in the United States, with 2.7 million square feet of operational space and 315 megawatts of power capacity. The multi-building campus can scale to 495 megawatts and accommodates enterprise, government, cloud, hyperscale, AI, and high-performance computing deployments under Switch’s Tier 5 Platinum standard.

The campus accommodates power densities of up to 55 kW per cabinet, allowing customers to consolidate high-density computing equipment into smaller footprints.

Switch reports an annual PUE of 1.28 for its Las Vegas facilities and provides a 100% power uptime guarantee. Modular data halls, redundant electrical systems, air and liquid cooling options, and continuous environmental monitoring protect mission-critical equipment.

Physical protection includes military-trained security personnel, a 24-hour security presence, controlled access points, surveillance systems, and roof structures rated for winds of up to 200 mph. Las Vegas provides low natural-disaster exposure, while Nevada offers a 2% sales and use tax and a 75% reduction in personal property tax for qualifying deployments.

The campus operates through Switch’s 100% renewable Green Power program and has achieved net-zero operational carbon emissions.

Its carrier-dense network provides connectivity to cloud and managed-service providers, with latency of 5 milliseconds to Los Angeles and a dedicated 7-millisecond fiber route to the Citadel Campus through the Switch Superloop.

This connection allows customers to distribute workloads between Las Vegas and Reno for active-active operations, disaster recovery, and geographic redundancy.

4. Google Data Center (Council Bluffs, Iowa, USA)

• Company: Google
• Location: Council Bluffs, Iowa, USA
• Power Capacity: Over 100 MW
• Area: 269,000 sq. meters (2.9 million sq. ft)
• Total Investment: Part of $20B+ Iowa investment
• CEO: Sundar Pichai
• Campus Scale: Four major buildings
• PUE: 1.08 to 1.11 (trailing 12-month average)
• Cooling System: Water-based evaporative towers; automated monitoring
• Renewable Energy Use: 90%+ hourly carbon-free; wind and solar agreements

Google’s Council Bluffs campus spans 2.9 million square feet across four major buildings. Operational since 2009, the site supports Google Search, Gmail, YouTube, Google Cloud, AI systems, storage platforms, and other digital services used worldwide.

Google has invested more than $20 billion in Iowa since selecting Council Bluffs in 2007. A further $7 billion commitment announced in 2025 covers continued expansion of the existing campus and construction of a new data center in Cedar Rapids.

The facility uses custom-built Google servers, high-speed internal networks, automated monitoring, layered physical security, and extensive fiber connections to other parts of Google’s global infrastructure. Water-based cooling systems remove heat from the computing equipment, while backup power and redundant electrical systems support continuous operations.

Google has paired the campus with major clean-energy purchases, including a 100-megawatt solar agreement. Iowa provides large industrial sites, reliable electricity, strong fiber routes, available land, and a central U.S. location for cloud, AI, and data-intensive workloads.

3. Harbin Data Center (Harbin, China)

• Company: China Mobile
• Location: Harbin, Heilongjiang Province, China
• Power Capacity: 200 MW
• Area: 663,000 sq. meters (7.1 million sq. ft)
• Cost: Not disclosed
• Operator: China Mobile (no publicly listed CEO for this facility)
• Rack Capacity: 50,000 racks across 52 modules
• PUE: Annual average PUE of 1.4 for phase one
• Cooling System: Heat-pipe rear-door; natural cold-source; heat recovery
• Tier Rating: ODCC 4A operational reliability (Phase 1)
• Rack Space: 10,252 commissioned; 50,000 planned

Located in northeastern China, the China Mobile Harbin Data Center covers 7.1 million square feet and serves as one of China Mobile’s major cloud, telecommunications, AI, government, and enterprise computing hubs.

The campus was designed with 50,000 data racks across 52 building modules, along with offices, teaching facilities, employee dormitories, and an exhibition hall. The 7.1 million-square-foot figure covers the full planned complex rather than dedicated data halls alone.

The development includes 18 data center buildings, 18 cooling stations, 11 management buildings, and 4 substations, backed by 12 billion yuan in planned investment. The first three construction phases provide capacity for more than 29,000 racks.

The first phase contains 96 modular computer rooms with rack configurations of 5 kW, 7 kW, and 10 kW. Power infrastructure includes 150 MVA of utility capacity, 46.4 MVA of diesel generation, 15 minutes of UPS runtime, and 72 hours of stored generator fuel. Natural cold-air cooling, chilled-water systems, heat exchangers, and heat recovery support a designed PUE below 1.4.

China Mobile’s Harbin Intelligent Computing Center entered service in 2024 with 18,000 AI accelerator cards and 6.6 exaFLOPS of computing power. The cluster supports large-model training, cloud platforms, government systems, telecommunications services, and high-performance enterprise workloads.

2. China Mobile Hohhot Data Center (Hohhot, China)

• Company: China Mobile
• Location: Hohhot, Inner Mongolia, China
• Power Capacity: 150+ MW
• Area: 715,000 sq. meters (7.7 million sq. ft)
• Cost: $1.9 billion
• Operator: China Mobile (no publicly listed CEO for this facility)
• Server Capacity: Up to 1.2 million servers
• PUE: 1.15 to 1.24 across various buildings
• Cooling System: Cabinet-level; indirect evaporative; immersion liquid cooling
• Renewable Energy Use: 77% green power (Q1 2024)
• Rack Space: 204,000 standard racks planned

China Mobile’s Hohhot campus is the world’s second-largest data center by reported area, spanning 7.7 million square feet. Announced in 2014, the development serves as a core node in China Mobile’s national cloud, telecommunications, data storage, and computing network.

The campus is planned with 18 data center buildings and capacity for 1.2 million servers. Its intelligent computing center delivers 6,700 petaflops of AI computing power, supporting large-model training and applications across government, finance, healthcare, transportation, education, and energy. More than 85% of its intelligent computing capacity uses domestically produced AI chips.

Inner Mongolia’s cold climate reduces mechanical cooling demand, while abundant renewable energy, available land, and high-capacity fiber connections support continued expansion. The campus forms part of China’s national computing network, transferring intensive cloud and AI workloads from eastern economic centers to the energy-rich Hohhot region.

The 7.7 million-square-foot figure covers the full campus plan, including data halls, power infrastructure, cooling systems, offices, and operational support facilities.

1. China Telecom Inner Mongolia Information Park (Hohhot, China)

• Company: China Telecom
• Location: Hohhot, Inner Mongolia, China
• Power Capacity: 150 MW
• Area: 994,062 sq. meters (10.7 million sq. ft)
• Cost: $2.6 to $3 billion
• CEO: Ke Ruiwen
• Rack and Server Capacity: 100k+ racks; 1M+ servers at full build-out
• PUE: 1.14 to 1.15 for newer facilities
• Cooling System: Natural cooling; indirect evaporative; cold-plate liquid cooling
• Tier Rating: ODCC AAAA green data center rating
• Renewable Energy Use: 95%+ green electricity including purchases (2026 report)
• Rack Space: 50k current racks; 100k+ planned

China Telecom’s Inner Mongolia Information Park is the world’s largest data center complex by reported area, spanning 10.7 million square feet in Hohhot. The 17.3 billion yuan development serves as a northern hub for China Telecom’s cloud, colocation, storage, telecommunications, and government computing services, with direct access to the company’s nationwide fiber network.

The master plan covers 42 data center buildings and 19 supporting buildings, with capacity for more than 100,000 racks and 1 million servers at full build-out. The complex includes data halls, power centers, offices, dormitories, logistics facilities, and operational support buildings, meaning the 10.7 million-square-foot figure represents the entire campus. Its customer base includes major technology companies, financial institutions, government agencies, and enterprises, with Tencent, Alibaba, and Baidu among the organizations connected to the park.

Inner Mongolia’s cold and dry climate reduces mechanical cooling demand, while abundant land and regional energy resources support large-scale expansion. The campus forms part of China’s national computing network, which moves data-intensive processing from eastern population centers to western regions with greater power and land availability.

List of the Largest Data Centers in the World by Compute Power

The largest data center in the world by compute power is xAI’s Colossus 2 near Memphis, Tennessee, holding an estimated 1,112,000 H100-equivalents of AI computing capacity.

Explore the full ranking to compare the world’s largest AI data centers by compute capacity, IT power, chips, operators, and expansion plans:

5. OpenAI Stargate Abilene

• Operator: Oracle
• Primary User: OpenAI
• Location: Abilene, Texas, USA
• Compute: Approximately 509,000 H100-equivalents
• IT Power: 421 MW
• Hardware: Nvidia GB200 rack systems with an estimated B200 and B300 GPU mix
• Estimated GPU Count: 201,600
• Estimated Capital Cost: $15.9 billion
• Cooling: Air-cooled OptiChill free-cooling chillers
• Status: Operational, partial build-out.

Oracle owns the AI hardware and provides the capacity to OpenAI at the first operational Stargate campus. Epoch AI estimates that four of the eight planned buildings are active, giving the site about half of its projected computing capacity. 

Buildings five through eight could raise the campus to approximately 1.02 million H100-equivalents, 843 MW of IT power, 403,200 GPUs, and $31.9 billion in deployed infrastructure by late 2026. The wider campus is designed for about 1.2 GW of total facility power, including cooling and electrical overhead.

4. Anthropic-Amazon New Carlisle, Project Rainier

• Operator: Amazon Web Services
• Primary User: Anthropic
• Location: New Carlisle, Indiana, USA
• Compute: Approximately 686,000 H100-equivalents
• IT Power: 910 MW
• Hardware: Amazon Trainium2 accelerators
• Estimated Accelerator Count: Approximately 1.05 million Trainium2 chips
• Estimated Capital Cost: $34.5 billion
• Cooling: Direct air-based cooling
• Status: Operational and expanding.

New Carlisle forms the largest known site within Project Rainier, Amazon’s distributed AI infrastructure program for training and deploying Anthropic’s Claude models. 

AWS publicly confirmed that nearly 500,000 Trainium2 chips were operating when Rainier launched and that the system supplied more than five times the compute Anthropic used for its previous models. 

Epoch AI estimates that buildings one through sixteen are operational, with buildings seventeen and eighteen nearing completion and another fourteen buildings planned. 

Full expansion could reach approximately 1.75 million H100-equivalents and 1.93 GW of IT power by early 2028.

3. Meta Prometheus

• Operator: Meta
• Primary User: Meta
• Location: New Albany, Ohio, USA
• Compute: Approximately 763,000 H100-equivalents
• IT Power: 631 MW
• Hardware: Nvidia B200 GPUs
• Estimated GPU Count: Approximately 302,000
• Estimated Capital Cost: $23.9 billion
• Status: Operational and expanding.

Prometheus combines permanent Meta data halls, leased colocation facilities, and rapidly deployed weatherproof structures across New Albany Business Park. 

The unconventional design allowed Meta to add five operational tent-style buildings faster than conventional data centers could be completed. Additional capacity comes from Vantage-operated facilities and other buildings distributed across the campus. 

Epoch AI projects Prometheus to reach approximately 1.03 million H100-equivalents, 854 MW of IT power, 408,500 B200 GPUs, and $32.4 billion in infrastructure by the fourth quarter of 2026. New on-site gas generation and battery systems are being developed to support this expansion.

2. Microsoft Fairwater Atlanta

• Operator: Microsoft
• Developer: QTS Data Centers
• Primary Users: Microsoft and OpenAI
• Location: Fayetteville, Georgia, USA, within the Atlanta metropolitan area
• Compute: Approximately 769,000 H100-equivalents
• IT Power: 636 MW
• Hardware: Nvidia B200 GPUs in GB200 NVL72 systems
• Estimated GPU Count: Approximately 304,300
• Estimated Capital Cost: $24.1 billion
• Status: Operational.

Fairwater Atlanta began operating in October 2025 and expanded to four operational buildings by June 2026. 

Its two-story design increases GPU density and shortens the distances between processors, memory, networking equipment, and cooling systems. GB200 NVL72 racks use closed-loop liquid cooling, while air-cooled chillers reject heat outside the buildings with minimal operational water consumption. 

Dedicated fiber connects Atlanta with Fairwater Wisconsin and future Microsoft campuses through the AI Wide Area Network, allowing geographically separate facilities to operate as one distributed AI supercomputer.

1. Colossus 2

• Operator: SpaceXAI, built through xAI
• Primary Users: SpaceXAI, Anthropic, and Cursor
• Location: Memphis, Tennessee, USA
• Compute: Approximately 1.11 million H100-equivalents
• IT Power: 946 MW
• Hardware: Nvidia GB200 and GB300 systems
• Estimated GPU Count: Approximately 440,000
• Estimated Capital Cost: $35.8 billion
• Cooling: Air-cooled chillers and dry air-cooled condensers
• Status: Operational.

Colossus 2 is currently the world’s largest operational AI data center by both estimated compute capacity and IT power. Epoch AI estimates that it contains approximately 110,000 B200-class GPUs and 330,000 B300-class GPUs, providing about 2.20 × 10²¹ eight-bit operations per second. 

The first major cluster entered service in October 2025, followed by successive GB300 expansions that raised the site to its present scale by July 2026. 

Its power system combines grid connections, battery storage, and natural-gas turbines located partly across the Mississippi state border. The campus supports xAI workloads and disclosed external users, including Anthropic and Cursor.

Last Verified: July 21, 2026. The compute, power, hardware-count, and capital-cost figures are estimates rather than audited operator disclosures. Epoch AI updated its data-center rankings on July 19, 2026, and notes that the record for single-campus AI compute has recently doubled about every seven months.

Read More Similar Posts: 

• 10 Largest AI Data Centers in the World
• 10 Largest Data Centers in the UK
• 10 Largest Underground Data Centers 
• 5 Largest Underwater Data Centers

What Is the Largest Data Center in the World by Megawatts?

xAI's Colossus 2, near Memphis, Tennessee, is the largest single-site data center in the world by power capacity, drawing approximately 946 megawatts. The site edges out Amazon's Project Rainier in New Carlisle, Indiana, which operates at around 910 megawatts and trains Anthropic's Claude models. Both figures come from Epoch AI's mid-2026 satellite tracking and continue to climb as the campuses ramp.

Colossus 2 succeeds Colossus 1, the original xAI facility built in Memphis in 2024, which draws roughly 300 megawatts and houses about 200,000 AI chips used to train the Grok models. The two sites operate together over high-bandwidth fiber, and xAI has stated plans to push the wider Colossus complex toward 2 gigawatts.

Meta’s Hyperion data center campus in Richland Parish, Louisiana, is planned to reach 5,000 megawatts, which would make it substantially larger. Hyperion remains under construction and does not qualify as the world’s largest operational data center.

What Is the Largest Data Center in the World by Square Footage?

As mentioned previously, China Telecom’s Inner Mongolia Information Park in Hohhot, China, is the largest data center in the world by square footage.

The operational campus covers approximately 10.7 million square feet, or 994,000 square meters, and has a reported power capacity of about 150 megawatts. The complex contains multiple data halls supporting cloud computing, telecommunications, colocation, and large-scale data processing services.

Which Company Has the Most Data Centers in the World?

Digital Realty has the most data centers in the world, operating more than 300 facilities across over 55 metropolitan areas. Its global portfolio serves more than 5,000 customers and includes colocation, hyperscale, cloud connectivity, and AI infrastructure. Digital Realty’s facility count exceeds Equinix, which operates more than 270 data centers worldwide.

How Big Is a Data Center?

A data center can range from a small facility with less than 5,000 square feet of computer-room space to a hyperscale campus covering millions of square feet. The U.S. Department of Energy classifies data centers below 5,000 square feet of computer floor space as small, while IBM cites an industry definition requiring a hyperscale facility to contain at least 5,000 servers and occupy at least 10,000 square feet.

Modern facilities are usually measured by electrical capacity rather than building area because two buildings of the same size may support very different amounts of computing equipment. Large data centers commonly operate at tens of megawatts, while Uptime Institute classifies projects exceeding 100 megawatts as giant data centers.

What Is an AI Data Center?

An AI data center is a type of data center specifically designed to handle the massive computational and storage demands of artificial intelligence workloads.

It’s a high-performance facility optimized to run AI models, process big data, and support machine learning operations at scale. These centers are packed with specialized hardware like GPUs, TPUs, and high-bandwidth interconnects—far beyond the standard CPUs used in traditional data centers.

What Makes It Different from a Regular Data Center?

Feature	Traditional Data Center	AI Data Center
Main Workload	Web hosting, storage, apps	Training and running AI/ML models
Hardware	Mostly CPUs	GPUs, TPUs, NVMe storage, high-speed fabric
Power Needs	Moderate	Extremely high due to GPU-heavy workloads
Cooling	Standard	Advanced (liquid cooling, immersion, etc.)
Network	General-purpose	Ultra-low latency, high throughput
Use Cases	Websites, databases, email	Chatbots, computer vision, big data analytics

Final Thoughts 

China and the United States lead the global data center race on different terms. China leads in reported facility size, while the U.S. dominates in data center count and AI power capacity.

China Telecom’s 10.7 million-square-foot complex in Inner Mongolia is the largest by floor area, followed by China Mobile’s 7.7 million-square-foot Hohhot campus. In the U.S., Switch’s Citadel campus in Nevada is the world’s largest colocation site, spanning 7.2 million square feet with up to 650 megawatts of capacity.

Measured by power, AI campuses now take the lead. xAI’s Colossus 2 near Memphis draws roughly 946 megawatts, while Amazon’s Project Rainier reaches about 910 megawatts.

India is expanding through Yotta NM1, an 820,000-square-foot Tier IV facility with 50 megawatts of capacity. Europe’s CWL1 campus in Wales covers 2 million square feet and operates at about 148 megawatts. Norway highlights the efficiency race through Green Mountain’s hydropower-powered facilities, including a former NATO mountain vault cooled by fjord water.

Data center growth now depends on two priorities: greater scale for AI workloads and greater efficiency through renewable power and advanced cooling. These facilities support cloud computing, AI, streaming, storage, and much of the modern digital economy.

Brightlio Delivers United States Colocation Services

Finding the right data center in the United States starts with Brightlio. As a telecommunications and colocation broker, we have data center partners across every major US market, ready to provide solutions matched to your needs and budget. Simply share your space, power, and connectivity requirements, and we'll provide colocation pricing at no charge.

Beyond colocation, we offer connectivity, unified communications, cloud, infrastructure hosting, disaster recovery, and advisory services. This full suite of offerings gives your enterprise a complete technology solution through a single partner.

Contact Brightlio now to get started!

Frequently Asked Questions (FAQs)

1. Who has the largest data center in the world?

The largest data center in the world by reported floor area is China Telecom's Inner Mongolia Information Park in Hohhot, China, at a stated 10.7 million square feet. It supports large-scale cloud, telecom, and government workloads.

2. Is Switch the largest data center in the world?

No. Switch's Citadel Campus in Nevada is the largest colocation data center campus in the world at 7.2 million square feet, but China Telecom's Inner Mongolia park holds the overall floor-area title.

3. What is the largest data center by power capacity?

By IT power capacity, the largest single-site facility is xAI's Colossus 2 near Memphis at roughly 946 megawatts. The largest concentration overall is the Inner Mongolia cluster, which exceeds 3,000 megawatts across multiple operators.

4. What is a Tier 1, 2, 3, or 4 data center?

These are classifications by the Uptime Institute that define data center reliability and infrastructure:
- Tier 1: Basic capacity, no redundancy, ~99.671% uptime
- Tier 2: Redundant components, ~99.741% uptime
- Tier 3: Concurrently maintainable, multiple power/cooling paths, ~99.982% uptime
- Tier 4: Fault-tolerant, fully redundant systems, ~99.995% uptime

5. Is there a Tier 5 data center?

While not officially recognized by any authority bodies, some providers use “Tier 5” as a marketing term to describe data centers with extreme security, energy efficiency, and 100% uptime. It’s not a standardized certification.

6. Why does China build such large data centers?

China invests in large-scale data centers to support national cloud infrastructure, AI training, government services, and data localization requirements. Northern regions such as Inner Mongolia and Heilongjiang offer cold climates and low-cost power that suit hyperscale operations.

7. Which countries have the most large data centers?

The United States and China host the most massive data centers. The United States leads in commercial, cloud, and AI infrastructure, while China focuses on state-backed hyperscale campuses. India is expanding quickly through operators such as Yotta.

8. How much power do the largest data centers consume?

The largest sites draw hundreds of megawatts. Switch's Citadel Campus is designed for up to 650 megawatts, while the newest AI campuses exceed 900 megawatts of IT load. Power efficiency and renewable sourcing are central design priorities.

9. Are the largest data centers colocation facilities?

Some are and some are not. Switch, QTS, Vantage, Digital Realty, and Yotta operate colocation facilities where businesses can lease space. Sites such as Google Council Bluffs, the NSA Utah Data Center, and the AI campuses are owner-operated and not available for colocation.

10. Will larger data centers be built in the future?

Yes. Demand for cloud services and AI compute continues to drive record-breaking projects. Meta's Hyperion campus in Louisiana targets around 5 gigawatts, and analysts expect new single-site records through 2027 and 2028.

11. What is the purpose of such large data centers?

Large data centers are used for cloud computing, data storage, content delivery, disaster recovery, and AI processing. Hyperscale centers serve millions of users and run critical services for tech giants and governments.

12. Are all large data centers privately owned?

No, while many large data centers are owned by private companies like Switch, Microsoft, and Google, others are government-backed, such as China Telecom’s center or the U.S. NSA's Utah Data Center.

13. How much power do the world’s largest data centers consume?

Large data centers can consume hundreds of megawatts. For example, the Citadel Campus can draw over 650 megawatts. Power efficiency and sustainability are major design priorities to reduce carbon impact.

14. Are large data centers environmentally sustainable?

Some large data centers use renewable energy sources such as solar, hydro, or wind. Switch's Citadel Campus runs entirely on renewable power, Apple's Mesa facility is backed by Arizona solar, and CWL1 in Wales draws on Welsh hydroelectric power. Cold-climate operators like Green Mountain in Norway push efficiency further, cooling sites with fjord water on 100% hydropower.