TL;DR
A Rice University and Harvard Business School study finds colocation datacenters concentrate in urban areas to be close to business customers and minimize latency, while hyperscale operators favor lower-density locations that offer cheaper power, land, and construction. Network capacity, water and power availability, and regional infrastructure still constrain where large campuses are sited.
What happened
Researchers from Rice University and Harvard Business School analyzed county-level data across the US to map how datacenter types pick locations. Their statistical work shows third-party colocation and hosting firms tend to set up in cities where business customers — financial firms and other latency-sensitive users — are concentrated and where regulatory and security needs argue for proximity. By contrast, major cloud providers build large hyperscale campuses in a smaller number of lower-density regions that offer lower electricity, land, and construction costs, enabling economies of scale. The study highlights that hyperscale sites nonetheless require robust network links and sufficient power and water. Northern Virginia is cited as a major example: it accounts for roughly 10–15% of global hyperscale capacity, credited to early investment in network infrastructure. The authors also suggest future growth may blend regional hyperscale hubs with targeted colocation facilities where low latency is critical.
Why it matters
- Location choices shape regional economic impacts, including job creation, tax revenue, and infrastructure demand.
- Energy and grid planning must adapt to clusters of large power-hungry facilities in lower-density areas.
- Network investment timing influences which regions become durable datacenter hubs.
- Businesses with latency-sensitive workloads will remain tied to urban colocation options, affecting procurement and architecture decisions.
Key facts
- Study authors: Tommy Pan Fang (Rice University) and Shane Greenstein (Harvard Business School).
- Analysis uses county-level data to model how population density, industry mix, and operating costs predict datacenter siting.
- Colocation and hosting providers prefer urban locations to be near customers and to meet low-latency, security, and compliance needs.
- Hyperscalers concentrate large campuses in fewer, lower-density regions to take advantage of lower electricity, land, and construction costs.
- Hyperscale operations benefit from economies of scale but still need adequate network infrastructure, power, and water.
- Northern Virginia hosts roughly 10–15% of global hyperscale capacity, supported by early network investment.
- The study suggests future expansion could pair major regional hyperscale hubs with strategic colocation sites for latency-sensitive services.
- In Europe, land constraints, high energy costs, and long grid connection lead to investment in secondary markets like Oslo and Madrid.
What to watch next
- Emergence of hybrid deployment patterns combining regional hyperscale hubs with targeted urban colocation sites.
- How AI-focused datacenter demand, which may be less sensitive to latency, changes the geographic mix of new builds.
- Expansion of secondary European markets (for example, Oslo and Madrid) as developers respond to local land, energy, and grid constraints.
Quick glossary
- Colocation (colo): Third-party datacenter facilities where multiple customers rent space, often close to business customers to reduce latency and meet compliance needs.
- Hyperscaler: Large cloud providers operating massive, often campus-style datacenters designed for scale, such as to run cloud services and AI workloads.
- Latency: The time delay in transmitting data between points on a network; lower latency is important for real-time applications.
- Economies of scale: Cost advantages gained by increasing the scale of operations, typically lowering per-unit costs for things like power and construction.
- Network infrastructure: Physical and logical components (fiber, switches, interconnection points) that carry and route data to and from datacenters.
Reader FAQ
Why do colocation providers choose cities?
They locate near business customers to offer low latency, meet security and compliance needs, and serve industries concentrated in urban areas.
Why do hyperscalers prefer lower-density locations?
Lower electricity, land, and construction costs in those areas make large-scale campuses more economical and support economies of scale.
Are hyperscalers abandoning urban datacenters entirely?
Not according to the study; it anticipates a hybrid pattern with regional hyperscale hubs complemented by colocation sites where latency matters.
Is Northern Virginia a significant datacenter hub?
Yes — the study estimates it holds about 10–15% of global hyperscale capacity, helped by early network investment.
ON-PREM Why colos are city slickers and hyperscalers are country bumpkins One wants customers next door, the other wants cheap power Dan Robinson Thu 8 Jan 2026 // 17:38 UTC Datacenter building decisions tend…
Sources
- Why colos are city slickers and hyperscalers are country bumpkins
- Why Population Density Drives Edge Computing ROI
- Rural America Colocation Real Estate
- Challenges in U.S. Hyperscale Data Center Development
Related posts
- Tamarind Bio (YC W24) hiring infrastructure engineer to scale model serving
- Practical takeaways from a 2025 review of Go: concurrency, types, errors
- How Human Activity and Climate Signals Shaped 2025 Satellite Views