60% of their cloud capacity is now hosted in data centers they own directly. The hyperscalers — Google, Amazon, Microsoft, Meta — have reversed their historical leasing strategy to become landowners. The utopia of an immaterial cloud collapses against the physical constraints of artificial intelligence.
This race for property ownership reveals how digital capitalism is becoming an affair of territorial rent. Specialized land reaches $4 million per acre in Northern Virginia, creating a new geography of economic power.
The Essentials
- Hyperscalers own 60% of their capacity in their own centers versus 40% in leased facilities, according to Synergy Research Group
- Specialized land for data centers reaches $4 million per acre in Northern Virginia
- Amazon, Google and Microsoft together represent 70% of global investments in physical cloud infrastructure
- This geographic concentration creates new territorial monopolies and threatens the balance of local ecosystems
The Return of Physical Ownership: A Strategic Shift
Digital giants are rediscovering the virtues of land ownership. According to Synergy Research Group, hyperscalers now own 60% of their total data center capacity, reversing the balance of power with third-party operators who historically represented the majority of their infrastructure.
This shift is explained by the explosion in computing power needs for AI. A data center optimized for artificial intelligence consumes between 50 and 100 megawatts, compared to 10 to 30 megawatts for a traditional center. This energy intensity makes leasing prohibitive and drives toward direct ownership.
Amazon Web Services leads this transformation with 200 data centers in direct ownership by late 2024. Google follows with 150 directly-owned installations, while Microsoft is accelerating with 80 new centers acquired in 2024. Meta concentrates its efforts on 30 ultra-high-performance mega-centers, each representing $500 million in investment.
Specialized Land Creates a New Real Estate Bubble
The rush for ownership is exploding land prices in strategic zones. In Northern Virginia, the global epicenter of data centers with 70% of American internet traffic, specialized land reaches $4 million per acre, ten times local agricultural prices.
This inflation is spreading to emerging zones. In Ohio, the arrival of Intel and hyperscalers has multiplied land prices around Columbus by five. In Texas, the Dallas region is recording acquisitions at $2.5 million per acre for sites of at least 100 hectares.
Implantation criteria are transforming certain rural territories into high-tech industrial zones. Access to three independent electrical grids, proximity to fiber infrastructure and availability of 200 megawatts become the new determinants of land value. Farmers in West Virginia are selling their land 50 times its agricultural value to specialized brokers.
Geographic Concentration Threatens Local Balances
This geography of digital property ownership creates new territorial monopolies. Three corridors concentrate 80% of global investments: the American East Coast from Washington to New York, the London-Amsterdam-Frankfurt triangle in Europe, and the Tokyo-Seoul-Singapore axis in Asia.
Northern Virginia illustrates the effects of this concentration. Loudoun County hosts 20% of the world’s internet traffic on 3,000 square kilometers. This density creates supply tensions: local electrical demand is increasing by 15% per year, forcing the construction of new power plants and the importation of electricity from Pennsylvania.
Natural hydrogen emerges from laboratories and redefines global energy geography, creating new energy supply opportunities for these electricity-hungry infrastructures.
AI Accelerates the Race to Control Physical Infrastructure
Generative artificial intelligence is transforming infrastructure needs. A model like GPT-4 requires 25,000 Nvidia H100 graphics chips, distributed across multiple centers to ensure redundancy. This architecture requires direct control of the supply chain, from building design to thermal management.
Hyperscalers are developing their own chips to optimize this vertical integration. Google deploys its TPU processors in 40 owned centers. Amazon installs its Graviton chips in 60 directly-controlled installations. This technical specialization makes relocation to external providers impossible.
Liquid cooling, necessary for new generations of AI processors, requires major architectural modifications. Microsoft is rebuilding 30 existing centers with integrated cooling circuits representing 40% of total cost. This technical complexity justifies direct ownership rather than adaptation of leased centers.
The New Digital Rent Transforms Technology Capitalism
This evolution marks the return of industrial capitalism in the digital sector. The GAFAM companies are reproducing the model of nineteenth-century railway companies: control physical infrastructure to monetize the services that pass through it. Amazon, Google and Microsoft now perceive dual rent: on the use of their cloud services AND on the ownership of underlying infrastructure.
This land strategy generates new revenues. Amazon leases part of its centers to third parties for $200 per kilowatt installed per month, representing 20% gross margin. Microsoft develops partnerships with local operators, retaining land ownership and leasing technical space.
German codetermination resists automation better than the American labor market, suggesting that governance models influence how tech companies integrate their infrastructure.
Territories Face a New Geopolitics of Cloud
This race for ownership is redrawing the balance of power between tech companies and public authorities. States are developing land attractiveness strategies: Ireland offers industrial land at reduced prices to attract Meta and Google. Singapore is building artificial islands dedicated to data centers.
France is launching its “Sovereign Cloud” plan with 15 specialized business zones to attract European investments against American domination. Germany is developing public-private partnerships allowing hyperscalers to acquire public land in exchange for local employment commitments.
This land geopolitics creates new dependencies. Countries without suitable infrastructure become dependent on regional centers. Africa depends 80% on European centers, generating 200-millisecond latencies that limit the development of local AI applications.
The hyperscalers are transforming their economic model by becoming owners of the infrastructure that carries the digital economy. This shift toward land rent questions the utopia of a decentralized and democratic cloud. It reveals how artificial intelligence, far from dematerializing the economy, makes it more dependent than ever on territorial control and physical resources.