A cloud region is a provider-defined geographic market where customers can deploy public-cloud workloads. It normally contains one or more data centers or availability zones connected by high-capacity networks. Availability zones are separate failure domains inside a region; they are not regions themselves. Local zones, edge locations, content-delivery points of presence, and ordinary data-center sites are different infrastructure layers.

Region geography matters because the physical location of cloud infrastructure affects architecture and risk. A nearby region can reduce latency. An in-country region can help an organization satisfy data-residency, privacy, and sector-specific requirements. Multiple regions can support disaster recovery, business continuity, and geographic redundancy. A broad footprint also lets multinational organizations enter local markets while retaining a common cloud operating model.

The issue is becoming more important as artificial intelligence moves into production. Training clusters need large concentrations of power, cooling, accelerators, and network capacity, while inference often needs to run close to users or regulated data. More regions can create more places to access cloud and AI services, but region count does not measure capacity, service breadth, GPU availability, price, or reliability. It is a measure of geographic reach, not a complete measure of cloud quality.

For this ranking, an active cloud region is a geographically identifiable, provider-designated public-cloud region, or a clearly documented equivalent, that was officially launched, operating, and available for customer deployment on 4 August 2026. A qualifying region must appear in the provider's infrastructure documentation, service catalog, pricing selector, or regional availability material as a distinct deployment boundary or region ID.

The provider universe is limited to independently operated public-cloud platforms that expose a unified customer-facing service. This rule excludes white-label networks whose nominal regions are operated and sold by separate local partners. It also avoids treating every data center owned or leased by a hosting company as a cloud region.

The count does not include:

Separately named regions in the same metropolitan area are counted only when the provider exposes each as its own deployable region identifier or service boundary. This treatment applies, for example, to Gcore's numbered locations and Akamai's expansion regions. Merely operating two data centers in the same city does not create two regions.

Commercial sovereign partitions are included when they are open to public-cloud customers, even if they use a separate operator or control plane. That is why the count includes Azure in China, AWS China, and the AWS European Sovereign Cloud, while excluding Azure Government, AWS GovCloud, and Oracle's government, defense, and dedicated realms. Active request-only disaster-recovery regions are included when eligible commercial customers can deploy there.

Ties are resolved in the following order: active availability zones, countries with active regions, geographic breadth, and most recent verified launch. Gcore, Vultr, and Akamai each have 33 counted regions. None publishes a comparable global availability-zone total for this footprint, so the second tie-breaker applies. Gcore spans 22 sovereign countries plus Hong Kong as a separately regulated market, Vultr spans 20 countries, and Akamai's counted core-compute regions span 15 countries.

Kamatera's 21 data centers and Latitude.sh's 25 deployable locations are not counted as provider-designated cloud regions. CloudSigma's 40-plus partner regions are excluded because they form a white-label partner network rather than one unified public cloud. CoreWeave qualifies as a public cloud, but its 18 General Access regions fall below tenth-ranked Tencent Cloud. Under a broader ranking of customer-selectable infrastructure sites, the order would differ; this article ranks public-cloud regions.

Microsoft Azure ranks first with 63 active commercial regions. The figure is reconstructed from two official lists: 57 entries in Microsoft's global Azure public-cloud list and six Azure in China regions operated by 21Vianet. The footprint is densest in the United States and Europe, but it also reaches Brazil, Chile, Mexico, South Africa, the United Arab Emirates, Qatar, Israel, India, Southeast and Northeast Asia, mainland China, Australia, and New Zealand.

India South Central in Hyderabad is the newest active region in the current official list. Microsoft's current geographies page describes it as open, and its launch update appeared at the end of July 2026. The region expands Microsoft's Indian footprint beyond Central India, South India, and West India and is designed with availability zones. Denmark East was another recent opening.

Azure also has the broadest named development pipeline in the ranking. Its public materials identify future regions in Thailand, a second Danish region, a second Malaysian region, Finland, Taiwan, Greece, Saudi Arabia, and Greater Atlanta. Microsoft says East US 3 is planned for early 2027 with availability zones. None of those announced regions is included in the active total.

Many Azure regions support multiple availability zones, but Microsoft does not publish a single stable global zone count. Zone capability and product availability must be checked for each region and service. That distinction matters for regulated databases, resilient applications, and AI workloads because an active region may not offer every feature or accelerator.

Azure serves enterprises, developers, governments, software companies, financial institutions, healthcare organizations, retailers, and AI builders. The regional map is especially useful to multinational customers that need local data residency or paired-region disaster recovery. However, the 63 is not identical to Microsoft's broader “70-plus regions” marketing language. It includes active request-only disaster-recovery regions and the commercial Azure in China cloud, but excludes Azure Government and unlaunched regions. Azure leads second-ranked Oracle Cloud Infrastructure by 18 regions.

Oracle Cloud Infrastructure, or OCI, has 45 active regions in the commercial realm documented by Oracle. Its footprint is unusually distributed: OCI operates two commercial regions in markets including Australia, Brazil, Canada, Chile, France, Italy, Japan, Mexico, Saudi Arabia, Singapore, South Korea, Spain, the United Arab Emirates, and the United Kingdom. It also reaches Colombia, South Africa, Morocco, Israel, India, Indonesia, Serbia, Sweden, Switzerland, and the Netherlands.

The latest active region is Morocco West in Casablanca, released on 20 February 2026, shortly after Malaysia West 2 in Kulai on 2 February 2026. Oracle has announced a second Moroccan site at Settat and a future region in Nairobi. Those projects would deepen OCI's African presence beyond Johannesburg and Casablanca rather than simply adding more capacity in established US and European hubs.

OCI calls its zone-level units availability domains. Five counted commercial regions—Frankfurt, London, Ashburn, Chicago, and Phoenix—have three availability domains; the other 40 have one each. That produces 55 availability domains. The design helps explain how Oracle reached many national markets quickly, but it also means most OCI regions do not provide the same in-region multi-zone topology as Google Cloud or AWS. Customers in single-domain regions may need a second region for facility-level disaster recovery.

OCI's infrastructure supports Oracle databases and business applications, general-purpose compute, regulated workloads, national digital programs, and AI clusters. Its regional strength is especially relevant to organizations already standardized on Oracle software or seeking in-country database residency.

Oracle sometimes uses a broader “50-plus public cloud regions” message that crosses realm or deployment-model boundaries. This ranking uses only the current commercial-realm table and excludes government, defense, dedicated, and isolated sovereign deployments. OCI is 18 regions behind Azure and two ahead of Google Cloud.

Google Cloud reports 43 regions and 130 zones across six continents on a locations page updated 23 July 2026. Its largest clusters are in the United States, Europe, and Asia-Pacific, but the network also includes Brazil, Chile, Mexico, South Africa, Qatar, Israel, Saudi Arabia, Australia, and New Zealand. Google connects these locations through its private backbone and delivers a portfolio centered on data platforms, Kubernetes, analytics, machine learning, and Vertex AI.

Bangkok became Google's 43rd cloud region on 21 January 2026. The three-zone region strengthened a Southeast Asian footprint that already included Singapore and Jakarta, while complementing multiple markets in Northeast Asia. Sweden, opened in March 2025, was the preceding launch. Google has also announced a future region in Türkiye, but it remained outside the active list on the measurement date.

Google's 130-zone disclosure is one of the clearest in the industry and averages roughly three zones per region. That consistency can support low-latency replication and failure-domain separation without crossing a national border. It does not mean every product or accelerator is available everywhere; specialized AI hardware and newer managed services can have a smaller regional footprint.

Google Cloud serves digital-native companies, retailers, financial institutions, manufacturers, governments, media businesses, and AI developers. Its region terminology maps cleanly to this methodology: regions, zones, network points of presence, and future locations are presented separately.

Google is two regions behind OCI but has more than twice as many disclosed zone-level failure domains. It is six regions ahead of normalized AWS. The contrast shows why a region ranking should not be interpreted as a resilience ranking: the number of regions and the number of zones inside them answer different questions.

AWS reports 39 active regions and 123 availability zones. This ranking subtracts the two US GovCloud regions because access is restricted to eligible government-related workloads, leaving 37 counted commercial or openly available sovereign regions. The AWS European Sovereign Cloud remains in scope because AWS says the Brandenburg region is generally available to all customers. The two commercial China regions are also included, although local partners operate them and customers use a separate partition.

The Brandenburg sovereign region became generally available on 14 January 2026, following the opening of Asia Pacific (New Zealand) on 1 September 2025. The launch illustrates how regional expansion is increasingly about operational sovereignty as well as distance. AWS has named Saudi Arabia and Chile as future full regions, with seven availability zones planned across the two. Announced sovereign Local Zones are excluded because a Local Zone is an extension of a parent region, not an independent region.

AWS says each region has at least three availability zones, and some have more. The official 123-zone total nevertheless covers all 39 active regions, including GovCloud. AWS does not publish a comparable subtotal for the 37 regions counted here, so none is reported.

AWS is especially dense in North America, Europe, and Asia-Pacific, with active regions in South America, Africa, and the Middle East as well. The platform serves startups, software companies, governments, financial institutions, industrial enterprises, retailers, and AI developers. Its service depth and zone topology are often stronger than a raw region count can show.

The government-only exclusion moves AWS below Google even though AWS's headline total is 39. On the same commercial-access basis used for Azure and OCI, AWS is six regions behind Google Cloud and three ahead of Huawei Cloud. The normalization is the main limitation readers should remember when comparing this rank with provider marketing.

Huawei Cloud reports 34 regions and 102 availability zones. Its center of gravity is mainland China and the wider Asia-Pacific market, but it also has a notable presence in Latin America, Africa, and the Middle East. The official page identifies two-region markets in Beijing and Mexico and says its infrastructure serves customers across Asia-Pacific, Latin America, Africa, Europe, and the Middle East.

The count rose from 33 regions and 96 zones at the end of 2024 to 34 regions and 101 zones at the end of 2025, then to 102 zones on the live page by 4 August 2026. Huawei's English-language materials do not clearly identify and date the 34th region. The latest individually documented full public-cloud launch located was the three-zone Philippines region in Manila on 3 December 2024.

Huawei's zone total suggests substantial multi-site redundancy, although services and topology vary by location. The provider also warns that partner cloud nodes are subject to updates. Its infrastructure serves telecom operators, public agencies, banks, manufacturers, retailers, and internet businesses, including customers adopting Huawei's data and AI stack.

Huawei's position is strongest in markets where its telecom relationships and local partnerships support cloud expansion, particularly China and several emerging economies. Procurement restrictions, regulatory requirements, and partner operating models can still affect who can use particular locations and which services are available.

Huawei is three regions behind AWS and one ahead of the three-way 33-region tie. Its headline total is official, but the lack of a fully enumerated English-language region list is a material transparency limitation. The count relies on Huawei's current infrastructure statement and its annual-report trend rather than a row-by-row public reconciliation.

Gcore ranks sixth with 33 active cloud regions. Its virtual-machine page says customers can deploy across more than 30 cloud regions, while the live cloud-pricing selector enumerates 33 current VM region entries. Those entries span North and South America, Europe, Africa, the Middle East, Asia, and Australia, giving Gcore one of the broadest country maps among the smaller public-cloud platforms.

Europe is Gcore's densest market. The live list includes multiple region identifiers in London, Luxembourg, São Paulo, and Sines, as well as locations such as Amsterdam, Darmstadt, Frankfurt, Helsinki, Paris, and Warsaw. It also extends to Almaty, Baku, Dubai, Hong Kong, Incheon, Istanbul, Johannesburg, Mumbai, Singapore, Sydney, Tashkent, Tokyo, Chicago, Manassas, and Santa Clara. The count treats numbered same-metro entries as separate only because Gcore exposes each one as a distinct region choice.

The latest publicly documented launch found was Sines-2 and Sines-3 in Portugal on 23 June 2025. Gcore described both as new cloud regions and made servers available for immediate deployment, including virtual machines, storage, GPU infrastructure, and other IaaS capabilities.

Gcore does not publish a comparable global availability-zone total. Its regions can also have different product catalogs: the presence of a VM price in a region does not guarantee that managed databases, object storage, serverless functions, or every GPU type are available there. Customers must check the service matrix and capacity for the specific region.

The platform targets developers, gaming and media companies, retailers, financial services, AI users, and organizations seeking regional data residency. Gcore's broader network marketing refers to more than 50 cloud locations and a much larger edge footprint, but this ranking uses only the 33 VM region entries. Gcore is one region behind Huawei. It wins the 33-region tie because its map covers 22 sovereign countries plus Hong Kong as a separately regulated market, ahead of Vultr's 20 countries and Akamai's 15.

Vultr operates 33 cloud data center regions across 20 countries. Its largest concentrations are in the United States and Europe, but it also reaches Brazil, Chile, Mexico, South Africa, India, Singapore, Indonesia, Japan, South Korea, Australia, and other Asia-Pacific markets. The footprint is designed around straightforward self-service deployment rather than the very large multi-zone campuses of the hyperscalers.

Milan opened on 19 May 2026 as Vultr's 33rd region and ninth in Europe. The launch expanded a European map that already included Amsterdam, Frankfurt, London, Madrid, Paris, Stockholm, Warsaw, and Manchester. Vultr had not named another unlaunched full region in any public materials.

Vultr uses “cloud data center region” for a customer-selectable deployment location. This article treats that term as equivalent to a public-cloud region because each location is exposed through the same platform and supports regional compute deployment. Vultr does not publish a globally standardized availability-zone count, and customers should not assume that every region contains multiple independently usable zones.

The platform serves developers, startups, SaaS companies, gaming workloads, media applications, and enterprises using virtual machines, bare metal, Kubernetes, storage, and selected GPU services. Product and capacity availability still vary by region, particularly for accelerated computing.

Vultr ties Gcore and Akamai at 33 regions. Because none of the three publishes a comparable global AZ figure, country coverage breaks the tie. Vultr's 20-country map places it behind Gcore's 22-country footprint plus Hong Kong but ahead of Akamai's 15-country core-compute footprint. Vultr is one region below Huawei and one rank above Akamai.

Akamai Cloud, built from the former Linode platform, has 33 active core-compute regions under this methodology. The main availability page lists 31 region IDs across North America, Europe, Asia, South America, and Oceania. Akamai's 2026 documentation adds Paris 2 and Washington 2 as distinct region IDs available to all customers, producing the 33-region total.

North America and Europe are the densest parts of the map. Akamai also operates core compute in Brazil, India, Indonesia, Japan, Singapore, and Australia. Paris 2 and Washington 2 added new capacity in existing metropolitan markets. They count separately because Akamai exposes each as its own region ID in the deployment interface, not merely as another facility inside an existing region.

Three legacy sites require special treatment. Akamai says new-resource creation in the original Jakarta, London, and Washington regions is disabled for new customers, while customers with pre-existing resources can continue deploying there. Those regions remain operational and customer-deployable for an existing customer group, so they are included. If Akamai fully retires them, its active total will fall unless new regions replace them.

Akamai does not expose a globally standardized availability-zone layer for core compute. Its much larger CDN and edge network is also excluded. The platform serves developers, SaaS companies, media and gaming workloads, retailers, and applications that combine centralized cloud compute with Akamai's delivery and security services.

Akamai loses the 33-region tie on the country criterion: its counted core regions span 15 countries, compared with Vultr's 20 and Gcore's 22 sovereign countries plus Hong Kong. It ranks one region ahead of Alibaba Cloud.

Alibaba Cloud lists 32 regions and 105 availability zones. Its greatest density is in mainland China, where it has more public-cloud regional endpoints than any other provider in this ranking. Internationally, Alibaba operates across East and Southeast Asia, Australia, Europe, the United States, the Middle East, and Mexico. It had no active full region in Africa or South America on the official list at the measurement date.

France became the newest active region on 17 June 2026, eight days after Alibaba opened Johor in Malaysia. The Paris launch expanded Alibaba's European presence, while Johor added a second Malaysian market alongside Kuala Lumpur. Alibaba's international expansion announcement also named first data-center markets in Brazil and the Netherlands. Those future sites are excluded until Alibaba confirms a public-cloud region launch and makes the regional endpoint available.

The 105-zone total indicates that many Alibaba regions contain multiple failure domains, although the topology and service catalog vary. The platform supports e-commerce, gaming, media, retail, manufacturing, databases, analytics, and AI, with a particularly strong position among organizations operating in or connecting to mainland China.

The main counting caveat is Alibaba's terminology. Its official total includes Nanjing, Fuzhou, and Wuhan as Local Regions with distinct region identifiers and regional cloud services. They are retained because Alibaba exposes them as region endpoints rather than as Local Zones attached to a parent region. Nanjing and Fuzhou were still supported but marked for closure. Excluding all three Local Regions would reduce the comparable total to 29, so readers should treat 32 as the provider's own broad region definition.

Alibaba is one region behind Akamai and nine ahead of Tencent Cloud. Its active count is well supported by the live location list, but future retirements of Local Regions could change the total even without a change in the broader physical footprint.

Tencent Cloud reports 23 regions and 66 availability zones. Its strongest presence is in mainland China and the wider Asian market. The international map includes Hong Kong, Singapore, Thailand, Indonesia, Japan, South Korea, Germany, the United States, Brazil, and Saudi Arabia. The regional network aligns with Tencent's strengths in gaming, media, communications, e-commerce, and consumer internet services.

The most recent active regional addition is Osaka. Tencent announced the new Japanese location in April 2025, and its May 2026 update identified Osaka among the additions that brought the footprint to 23 regions and 66 zones. The live infrastructure page now lists Osaka with one availability zone alongside two zones in Tokyo. Riyadh and added capacity in Frankfurt were also recent expansions, but Osaka is the newest distinct active regional market reflected in the current count.

Tencent has also announced a Mexico cloud region with two availability zones. The announcement was made on 15 June 2026. Mexico is a future region and is not included in the 23-region total.

Tencent's 66-zone total averages nearly three zones per region, but individual regions range from one zone to much larger multi-zone deployments in China. Its own documentation defines a region as a geographically isolated area containing one or more zones, which aligns well with this methodology. Product availability, account access, and network arrangements still vary between mainland China and international markets.

Tencent is nine regions behind Alibaba and ranks five regions ahead of CoreWeave's 18 General Access regions. It serves Chinese companies expanding overseas and international organizations reaching Asian users, but its European and American region density remains lower than that of the providers ranked above it.

The ranking reveals several expansion models. Azure, OCI, Google Cloud, and AWS operate global hyperscale networks, but they make different trade-offs between region count and the number of failure domains inside each region. Huawei, Alibaba, and Tencent combine large Asian footprints with targeted international expansion. Gcore, Vultr, and Akamai distribute smaller customer-deployable regions across many markets, often with less standardized zone disclosure.

Provider headlines are not directly comparable. Azure's “70-plus” language is broader than its reconciled commercial lists. AWS's 39 includes GovCloud. Oracle's broader totals cross realm and deployment boundaries. Huawei publishes a headline without a complete English-language enumeration. Gcore's network and cloud-location counts are broader than its VM region list. Akamai's edge network is far larger than its core-compute footprint. Alibaba includes three Local Regions. Normalization materially changes the order.

Region count should therefore be the start of infrastructure due diligence, not the conclusion. Buyers must also compare service availability, zone topology, capacity, accelerator inventory, inter-region networking, egress costs, certifications, operational independence, and the legal entity responsible for each location. A nominally local region has limited value if the required database, GPU, security control, or support model is unavailable there.

Future growth is moving toward markets that have historically depended on distant hubs. The announced pipelines include Chile, Mexico, Brazil, Saudi Arabia, Kenya, Türkiye, Greece, Thailand, Taiwan, the Netherlands, and additional sites in Morocco, Denmark, Malaysia, Finland, and the United States. New regions can reduce latency and create local cloud ecosystems, but their economics depend on power availability, fiber connectivity, construction capacity, regulation, and sustained demand.

Artificial intelligence is reshaping site selection. Training infrastructure favors large campuses with dense power, advanced cooling, accelerators, and high-speed networking. Inference creates pressure to distribute capable compute closer to users and data. The result may be a two-layer map: a smaller number of very large AI training hubs and a wider set of regional inference and general-cloud locations.

Data sovereignty is the second major force. The AWS European Sovereign Cloud, Azure's national expansion program, Oracle's Moroccan build-out, and new in-country regions across Asia and the Middle East show that customers increasingly ask who operates infrastructure, which jurisdiction governs support access, and whether workloads can remain in-country. Sovereignty can create new public regions, but it can also produce restricted or partner-operated clouds that should not automatically be added to commercial-region totals.

Latency and regulation are reinforcing each other. Governments want control over sensitive data, while businesses and consumers expect real-time cloud and AI services. Providers that can combine a local region, multiple failure domains, enough capacity, a useful product catalog, and compliant operations will be best positioned in underserved markets. The regional map will keep expanding, but the openness, resilience, and service depth of each region will matter more than the headline number alone.