In the past few decades, data centers have evolved from simple computer rooms in the 1960s to massive, purpose-built facilities powering today’s digital economy. Historically owned and operated by large corporations for their own use, dedicated hosting facilities emerged where multiple companies could house their equipment in shared spaces through a ‘colocation’ model as internet adoption grew in the 1990s. The 2000s saw the rise of wholesale colocation data centers and hyperscale facilities driven by the growth of cloud computing and, more recently, by AI computing.

As the sector expands, investing in data centers has also become a lucrative business, attracting financial investors. While Private Equity remains the main player, other types of investors are coming to the scene, including through Real Estate Investment Trusts (REITs) and venture capital investment.

Key Players

As it stands, three types of players exist in today’s data center sector.

Hyperscalers

These are companies that develop, own, and operate data centers and provide service on top. While telecommunication companies were the first generation of ‘hyperscalers’ and shaped the initial data center industry, today’s hyperscalers are mainly large cloud providers including Amazon Web Services (AWS), Microsoft Azure, and Google Cloud. More recently, the new computing demand from AI training has been driving a new wave of expansion from both existing hyperscalers as well as new players such as Meta, which historically has a relatively small footprint in data centers.

Specialized Operators

A large chunk of the data centers is owned and developed by specialized operators under a business model called ‘colocation.’ Equinix, founded in 1998, has grown to become the world’s largest colocation provider with a global footprint of interconnection-focused facilities. Digital Realty stands as one of the largest global providers, offering a comprehensive portfolio that spans both wholesale and colocation services across major markets. CyrusOne has established itself as a major player specializing in enterprise-class data center solutions. CoreSite completes this core group of traditional operators with its specialization in network-dense colocation facilities that focus on creating rich ecosystems of connectivity options for customers.

Financial Investors

As the sector grows, data center investment has also evolved from a niche real estate strategy to a mainstream investment strategy. Private Equity (PE) has been the largest financial player, usually through infrastructure investing. The invention of Real Estate Investment Trusts (REITs), a liquid, public-trading vehicle, unlocks additional capital from both retail investors and mainstream institutional investors demanding high liquidity.

  • Private Equity investors have brought significant capital and sophisticated investment strategies to this rapidly growing infrastructure segment. The two most common strategies include directly investing in the underlying asset through an infrastructure strategy or investing in Data Center operators, such as the JV formed between Blackstone and QTS.
  • Structuring the assets as REITs attracts additional investors by lowering the capital commitment and providing more liquidity. By packing the underlying real estate assets into REITs, operators can also unlock additional funding and increase revenues. Digital Realty Trust, for example, offered by Digital Realty, is held by major asset owners and asset managers, including Norges Bank, JPM, Blackrock, etc

Data Center Development Process and Business Model

Development Process

Data center development follows a progressive approach, with costs and capabilities increasing at each stage of development. Depending on the development process, business models vary.

  • At the most basic level, a Shell Data Center consists of just the building structure without power or cooling infrastructure. This bare building envelope with basic utilities typically takes 6-12 months to construct and usually represents 30-40% of the total development cost. This initial stage provides the foundation upon which more specialized infrastructure can be built, allowing for strategic phasing of capital deployment.
  • Moving up the value chain, a Powered Shell includes the building with power infrastructure but minimal IT-specific fit-out. This stage includes electrical substations and primary power distribution systems, with capacity secured but not fully distributed throughout the building. Powered shells are particularly attractive for tenants wanting customization flexibility while benefiting from having the most capital-intensive utility infrastructure already in place. This intermediate stage reduces tenant lead time while preserving options for customized deployments.
  • The Warm Shell represents another step toward readiness, featuring power plus some cooling infrastructure. These facilities include basic cooling systems and raised floors, with partial infrastructure deployment that significantly reduces tenant fit-out time. This model effectively balances flexibility with faster deployment time, making it suitable for customers with defined requirements but some specific customization needs. The pre-deployment of key infrastructure systems at this stage dramatically accelerates the time to operational status.
  • A Turn-Key Data Center represents a fully built facility ready for IT equipment installation without further development work. These centers feature complete mechanical, electrical, and plumbing systems, along with comprehensive security systems, monitoring capabilities, and network connectivity. Turn-key facilities are ready for immediate equipment deployment, allowing customers to focus exclusively on their IT infrastructure rather than facility development. This complete solution comes at a premium but eliminates facility development risk and time delays.
  • The final and most comprehensive stage is the Operational Data Center, which is a fully staffed and managed facility. These centers feature on-site technical and security personnel, ongoing maintenance and monitoring services, and comprehensive customer support and service delivery infrastructure. This full-service approach transforms the data center from a physical asset into an operational service, with the associated costs and complexity managed by specialized providers with expertise in facility operations.

Business Model

Depending on the player type and development stage, at least three business models exist when operating a data center.

Colocation

Most special operators operate under a Colocation (often shortened to “colo”) business model by developing data centers up to a certain stage and renting physical space to their customers for additional customization.

  • Wholesale colocation, for example, targets larger enterprises requiring substantial capacity, security, and customization, typically leasing entire data halls or buildings with deployments exceeding 1MW of power. The offerings often align with earlier stages of data center development, such as powered shell or warm shell configurations, allowing customized facilities to meet special requirements such as transmission bandwidth and diversity of on-net carriers. The lease is usually long-term, extending beyond 10 years.
  • Retail colocation, on the other hand, primarily serves small to medium-sized businesses demanding smaller space and less customization. The facilities are typically fully built out as turn-key environments, with complete power distribution, cooling systems, raised floors, and security infrastructure already in place, with customers bringing in their own IT equipment and operating under minimal lead time. The pricing structure is similar to apartment renting, usually including fixed monthly charges for space, such as racks, cabinets, or cages, and variable charges for power consumption and connectivity services.
Service Provider Model

While the Colocation model is a Real Estate business model, most hyperscalers operate under the Service Provider model. The development costs of the data center essentially become the Opex of operating the cloud infrastructure, and the revenue of a data center is the additional service; hence, billing can be provided by the associated cloud service. AI data centers, mostly strategic investments at this point, have also been adopting this model in their thinking.

Bare Metal as a Service (BMaaS) Model

Another interesting business model that lies between the two types is the Bare Metal as a Service (BMaaS) Model, which offers dedicated physical servers provisioned on-demand in addition to just the cloud. Revenue in this model combines subscription fees based on hardware specifications with data center facilities charges, including storage, networking, security, and usage-based bandwidth. BMaaS targets customers with performance-sensitive workloads and high-compute applications who need cloud-like flexibility but require dedicated hardware performance. Notable providers include IBM Cloud Bare Metal, Equinix Metal, OVHcloud, Leaseweb, and Hetzner Dedicated Servers.

Bottlenecks and Opportunities in the Era of Energy Transition

Location Constraints

The data center industry faces significant bottlenecks in development as it navigates the complexities of the ongoing energy transition.

  • Securing cheap energy access has replaced land rights as perhaps the most pressing constraint in location siting. Projects in key markets are facing delays or cancellations due to insufficient power infrastructure. As a result, locations with abundant cheap power access are attracting disproportionate investment.
  • Another key consideration is access to fiber connectivity and security. Data centers require robust fiber infrastructure with redundant pathways to ensure reliability, as well as sufficient physical and cyber security to ensure data safety.

Alternative solutions are being developed to unlock more lands.

  • On the energy access side, many are looking at offgrid or hybrid solutions to supply power directly on-site through a combination of renewable energy, energy storage, and baseload power ranging from mature technologies such as solar, wind and batteries, to emerging solutions such as hydrogen and nuclear.
  • On the fiber access side, as primary fiber routes become congested, developers are exploring alternatives to traditional fiber connectivity, including satellite-based solutions for remote locations, microwave and millimeter wave technology for low-latency applications, and even early-stage quantum communications networks that promise unprecedented security capabilities.
Design and Construct Energy-Efficient Data Centers

The imperative to build more sustainable facilities has spawned innovation in both materials and design methodologies.

Sustainable materials represent a major opportunity for reducing the environmental footprint of data centers beyond just operational energy use. Low-carbon concrete and steel alternatives can significantly reduce the embodied carbon in these massive structures, while modular construction approaches minimize material waste through precision manufacturing. The industry is also embracing recyclable and reusable building components that support circular economy principles, reducing life cycle impact.

Once operational, data centers face ongoing challenges in maintaining efficiency while meeting ever-increasing density and performance requirements. Energy-efficient computing and transmission, as well as more effective cooling, are critical to address such challenges. Advanced AI chips and photonics systems are expected to enhance computing and transmission with less energy. Advanced cooling, such as liquid cooling, direct-to-chip cooling, and dynamic cooling, which adjusts cooling based on workload using predictive algorithms, holds promise to reduce cooling-related energy consumption.

Furthermore, digital twin technology and specialized software tools have revolutionized data center design and operation. Software that combines simulation with energy modeling allows users to simulate thousands of scenarios and optimize for energy-efficiency design before construction begins. Digital twins of the operating facilities that collect data from networks of IoT sensors and train AI models for predictive analytics increase operating efficiency by preemptively identifying potential failures and minimizing unplanned downtime.

Mega or Small? And where will they go?

The data center landscape is evolving toward a more diverse ecosystem of facilities optimized for specific roles within the digital infrastructure. In the mega versus small dichotomy, hyperscale facilities exceeding 100 MW of power capacity are becoming increasingly common, housing massive cloud operations and providing economies of scale for standardized workloads. On the other hand, edge data centers under 1 MW are emerging to support latency-sensitive applications that require processing closer to end users.

The geographic distribution of data centers is also diversifying beyond traditional hub locations. Urban micro data centers are emerging to serve ultra-low latency applications in densely populated areas, often repurposing existing buildings in central business districts to place computing resources closer to users. At the opposite end of the spectrum, remote facilities are leveraging abundant renewable energy and natural cooling opportunities in locations far from population centers, connected via high-capacity fiber networks.

The divergence is driven by the two directions of computing:

  • Centralized computing for economy of scale is usually required for cloud infrastructure or AI computing.
  • Distributed computing, including edge computing, is used to process data generated for lower latency, usually preferred for consumer applications, including EVs, and likely more in the future, AR/VR, and even robotics.

While we are still early on in many of these new technology trends and are at a stage of developing the infrastructure layer, it makes sense that the new data centers being developed these days are largely at remote locations and of mega-scale, sometimes through a data center campus. As we move on to the application layer, where latency rather than computing power becomes the bottleneck, I foresee smaller data centers taking over a larger market share, especially as distributed computing technologies, including resource sharing and parallel processing, mature.

From Data Computing to Data Storage & Security?

The focus of data infrastructure is shifting beyond raw computing power to encompass the entire data lifecycle, with particular emphasis on storage and security. Data lifecycle management is evolving into a specialized service segment with dedicated infrastructure for different phases from ingest to archival or deletion. Cold storage facilities for archival data are growing in importance as organizations seek to preserve expanding data assets while minimizing active storage costs. The looming threat of quantum computing to current encryption standards is driving security infrastructure upgrades, with quantum-resistant encryption becoming a priority for future-proofing sensitive data storage. The next decade will likely see data centers evolve from primarily computing-focused facilities to comprehensive data management hubs addressing the full lifecycle of data, demanding new specialized infrastructure, innovative cooling and power solutions, and increasingly sophisticated security measures to protect against emerging threats.

From Today to Tomorrow

As data continues to grow as the world’s most valuable resource, the humble data center - once merely a “computer room” - will continue its transformation into one of the most critical components of global infrastructure, underpinning everything from everyday conveniences to groundbreaking innovations in AI, quantum computing, and beyond.

For investors and industry participants, the most significant opportunities will not lie in expanding raw capacity but in developing specialized solutions for specific workloads, focusing on efficiency practices, and creating the flexible infrastructure capable of supporting future technologies.