WHO IS DRIVING HYPERSCALE DATA CENTER DEMAND? Good Neighbor Data Audio Briefing — Episode 2 Published July 2026 This episode is educational and AI-narrated. It is not legal, engineering, utility, environmental, fiscal, investment, or planning advice. Figures and plans are current as of July 13, 2026, and announced projects can change. When people say “hyperscale data center demand,” they are often talking about a surprisingly small group of companies. Amazon Web Services. Microsoft Azure. Google Cloud. Meta. Oracle. These companies operate computing infrastructure at a scale that can reshape regional power planning, land markets, water discussions, construction pipelines, and local development policy. They are not identical. AWS, Azure, Google Cloud, and Oracle sell cloud capacity to outside customers. Meta primarily builds infrastructure for its own platforms and artificial intelligence systems. But all five deploy enormous fleets of servers, storage, networking, and increasingly, specialized artificial intelligence hardware. And all five use two paths to grow. They build and operate many of their own data center campuses. They also lease enormous blocks of wholesale capacity from third-party developers when they need to enter a market or add power faster than they can deliver a campus themselves. That build-versus-lease distinction matters. It explains why a data center proposed by an unfamiliar developer may ultimately be driven by one of the largest technology companies in the world. Start with the scale. Synergy Research Group counted 1,360 large data centers operated by hyperscale companies at the end of 2025. Those facilities represented 48 percent of worldwide data center capacity. Almost 60 percent of hyperscale capacity was in facilities the operators built and owned. The balance was leased. So the most accurate statement is not that hyperscalers only build their own facilities. It is that owned campuses remain the majority, while leased capacity is a central part of the operating model. Synergy expects hyperscale operators to represent 67 percent of worldwide data center capacity by 2031. It also tracks almost 800 future hyperscale facilities in the pipeline. Artificial intelligence is accelerating that growth, but the physical expansion began much earlier with search, ecommerce, social media, streaming, enterprise software, and public cloud computing. Amazon Web Services is the clearest place to begin. Amazon launched AWS in 2006 after learning how difficult it was to provision infrastructure for its own ecommerce business. The first major service, Amazon S3, offered storage on demand. A few months later, Amazon EC2 offered computing capacity on demand. The starting point was modest by present standards. AWS says S3 launched with roughly one petabyte of capacity across about 400 storage nodes in 15 racks spanning three data centers. Early EC2 offered a single instance type in one availability zone. The model changed the industry. Instead of buying servers and waiting for a facility, a developer could rent infrastructure through software. But the cloud still had to exist somewhere. As AWS expanded, it organized physical infrastructure into geographic regions. Each region contains multiple availability zones: separate groups of data centers with independent power, cooling, and networking. As of July 2026, AWS reports 39 launched regions and 123 availability zones, with additional regions announced for Saudi Arabia and Chile. It operates a global network with nearly 20 million kilometers of terrestrial and subsea fiber. AWS’s future plans are centered on both global cloud growth and artificial intelligence. Amazon has said it expects approximately 200 billion dollars of capital expenditures in 2026, much of it connected to AWS capacity that will be monetized in later years. That number should not be confused with spending on buildings alone. It includes land, power infrastructure, servers, networking equipment, chips, and other assets. AWS is also making large regional commitments, including a public plan to invest 20 billion Australian dollars in Australian data center infrastructure. Microsoft followed a different path into public cloud. The company already operated large infrastructure for services such as Hotmail and MSN. It announced Windows Azure in the late 2000s and made Windows Azure and SQL Azure generally available in February 2010 across 21 countries. Azure grew from a platform for Microsoft-oriented applications into one of the world’s largest cloud infrastructure systems. Microsoft now reports more than 80 Azure regions and more than 500 data centers worldwide. A region can contain several separate facilities, and a single facility can support many services, so region and data center counts should never be treated as interchangeable. Microsoft’s current expansion is being driven by customer cloud demand and by its own artificial intelligence products. The company has repeatedly said demand exceeds available supply. It expects roughly 190 billion dollars in capital expenditures during calendar 2026, including GPUs, CPUs, networking, long-lived facilities, and finance leases. Those finance leases are important. They show how even a company with one of the largest self-built fleets in the world uses leased campuses and capacity to expand. At the same time, Microsoft is developing giant owned or controlled projects. Its Fairwater campus in Wisconsin is planned to scale to two gigawatts. A new campus announced for Pecos, Texas, is also planned to add approximately two gigawatts over a five-to-seven-year buildout. These are not ordinary industrial loads. They require long-range planning for generation, transmission, substations, backup systems, construction labor, and community infrastructure. Google’s infrastructure story begins before Google Cloud existed. In 1999, Google’s first data center footprint was a tiny leased cage inside an Exodus facility in Santa Clara. It held roughly 30 personal computers on shelves. As Search, Gmail, Maps, and YouTube grew, Google moved into purpose-built campuses, designed its own servers and networking systems, and built one of the world’s largest private networks. Google entered the cloud platform market with services such as App Engine in 2008 and later expanded into infrastructure through Compute Engine. Today, the same physical system supports consumer products, Google Cloud customers, and Google’s own artificial intelligence models. Google reported 42 cloud regions and 127 zones in April 2025. By 2026, it described a network connecting 43 cloud regions and spanning more than 10 million kilometers of terrestrial and subsea fiber. The most revealing part of Google’s future strategy is not simply the number of buildings. It is the way Google links buildings together. The company says artificial intelligence demand can exceed the space and power available at any single facility. Its response is to locate data centers near viable energy resources, then use high-capacity networks to distribute workloads across campuses. In Google’s phrase, the continent becomes the data center. That means future hyperscale development may function less like a collection of isolated buildings and more like a networked regional supercomputer. Power availability can determine where compute is placed. Fiber determines how those locations operate as one system. Meta is different because it is not primarily selling public cloud infrastructure. Its data centers support Facebook, Instagram, WhatsApp, advertising systems, recommendation engines, and Meta’s artificial intelligence work. Yet in physical terms, Meta is unquestionably a hyperscale operator. Meta broke ground in 2010 on its first custom-built data center in Prineville, Oregon. That campus became a test bed for efficient server, cooling, and facility designs, many of which Meta shared through the Open Compute Project. The fleet has since expanded globally. In July 2026, Meta announced that a planned one-gigawatt data center in Sturgeon County, Alberta, would become its first data center in Canada and the thirty-third facility in its global fleet. Meta said the project represented more than 13 billion Canadian dollars of investment. But Meta also provides one of the clearest current examples of the lease model. In June 2026, Meta announced an agreement with Reliance Industries for a 168-megawatt artificial-intelligence-enabled data center in Jamnagar, India. Reliance will build the facility. Meta will lease it, with options to scale. That is hyperscale wholesale demand in plain language: a third party delivers the campus and power capacity; the hyperscaler becomes the long-term tenant and installs or operates the computing systems needed for its business. Oracle has the longest enterprise technology history in this group, but its newest data center expansion is tied to Oracle Cloud Infrastructure and large artificial intelligence contracts. Oracle launched its broad cloud platform in 2012 and later rebuilt its infrastructure offering around what it called Generation Two Cloud Infrastructure. Oracle’s footprint is smaller than AWS, Azure, or Google Cloud, but its expansion rate and project scale have changed rapidly. In January 2026, Oracle said it had 147 active data centers worldwide and 64 more on the way. It also described artificial intelligence projects with OpenAI at two campuses in Texas and at sites in New Mexico, Wisconsin, and Michigan. Oracle’s Project Jupiter in southern New Mexico is another important lease example. Oracle identifies itself as the tenant. Development partners are delivering the campus, and Oracle plans to deploy cloud infrastructure there for OpenAI. Oracle guided to 50 billion dollars of capital expenditures for fiscal 2026 and announced a major financing plan to add capacity for contracted cloud customers. It has also said it expects to deliver dozens of additional multicloud data centers inside or alongside the infrastructure of AWS, Google, and Microsoft. Oracle therefore illustrates two overlapping trends: hyperscalers leasing from developers, and hyperscalers interconnecting with one another because customers want services and data to move across multiple clouds. What does all of this mean when a new project appears before a community? First, identify every important party. Who owns the land? Who is developing the buildings? Who is financing construction? Who will own the utility interconnection? Who is the expected tenant? Who will operate the servers? Who is responsible for water, noise, backup generation, road improvements, emergency response, and eventual closure? The developer at the public meeting may not be the company creating the long-term demand. Second, review the full buildout, not only the first building. A hyperscale campus may be developed in phases over many years. A lease may include options for the tenant to take more capacity. A proposal that begins with tens or hundreds of megawatts may sit inside a much larger land, power, or infrastructure plan. Third, separate announcements from operating facts. A future region is not yet an operating region. A planned gigawatt is not a current gigawatt. A capital expenditure forecast is not a construction budget. And a renewable energy contract does not, by itself, explain hourly grid conditions, transmission needs, or who pays for local upgrades. Finally, understand why the demand is moving so quickly. Building a major owned campus can take years. Developers must secure land, power, permits, equipment, financing, and construction labor. Wholesale leasing can shorten that path because a specialist developer may already control a powered site or be able to deliver a built-to-suit facility. Hyperscalers use owned facilities for long-term control, standardization, and scale. They use leases for speed, flexibility, market entry, and access to scarce power. Artificial intelligence increases the value of both strategies because demand is growing faster than many companies can build. So when people talk about hyperscale data center demand, this is who they mean. A small number of companies with global digital platforms, enormous capital budgets, and rapidly growing computing requirements. They are building campuses measured in hundreds of megawatts and, increasingly, gigawatts. They are also leasing immense amounts of third-party capacity when the market can deliver it faster. For communities, the company name is only the beginning. The real work is to understand the complete chain of responsibility—and to make sure that speed and scale do not outrun public evidence, infrastructure planning, or measurable commitments. This has been a Good Neighbor Data audio briefing. For source links, transcripts, and practical review guides, visit goodneighbordata.com slash responsible-data-centers. SELECTED SOURCES Synergy Research Group — Hyperscale Operators to Account for 67% of All Data Center Capacity by 2031 https://www.srgresearch.com/articles/hyperscale-operators-to-account-for-67-of-all-data-center-capacity-by-2031 AWS — Our Origins https://aws.amazon.com/about-aws/our-origins/ AWS — Global Infrastructure https://aws.amazon.com/about-aws/global-infrastructure/ Amazon — 2025 Letter to Shareholders https://www.aboutamazon.com/news/company-news/amazon-ceo-andy-jassy-2025-letter-to-shareholders Microsoft — Windows Azure General Availability https://blogs.microsoft.com/blog/2010/02/01/windows-azure-general-availability/ Microsoft Datacenters https://datacenters.microsoft.com/ Microsoft — Fiscal Year 2026 Third Quarter Earnings Conference Call https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q3 Google Cloud — Google’s AI-Powered Next-Generation Global Network https://cloud.google.com/blog/products/networking/google-global-network-principles-and-innovations Google Cloud — Data Center and Global Networks Built for the AI Era https://cloud.google.com/blog/products/networking/data-center-and-global-networks-built-for-ai-era Meta — Prineville Data Center Expansion and History https://datacenters.atmeta.com/2021/03/facebooks-prineville-data-center-is-growing-again/ Meta — First Data Center in Canada https://about.fb.com/news/2026/07/breaking-ground-on-metas-first-data-center-in-canada/ Meta — Leased AI-Enabled Infrastructure in India https://datacenters.atmeta.com/2026/06/bringing-ai-enabled-infrastructure-to-india/ Oracle — AI Infrastructure in 2026 https://www.oracle.com/news/announcement/blog/oracle-ai-infrastructure-in-2026-and-our-commitment-to-local-communities-2026-01-26/ Oracle — Project Jupiter Tenant Announcement https://www.oracle.com/news/announcement/blog/oracle-advances-american-ai-innovation-in-new-mexico-2026-01-23/ Oracle — Fiscal Year 2026 Third Quarter Results https://oracle.com/news/announcement/q3fy26-earnings-release-2026-03-10/