The data center sector has become one of the most heavily invested, and the most complex, in the world. Global data center investment is running into hundreds of billions of dollars annually, individual campuses are targeting power requirements equivalent to those of major cities, and the regulatory and geopolitical environment surrounding the sector is evolving at a pace that challenges even the most sophisticated market participants.
In this report we examine the data center build-out from seven angles: joint venture structures, geopolitics and AI regulation, financing, cybersecurity, behind-the-meter power, GPUaaS, and environmental, social and governance (ESG). Within these dynamics, several common themes emerge.
Power has become the binding constraint on the sector’s growth. Capital requirements have outstripped the capacity of any single party to fund alone. Regulatory complexity, from export controls to foreign investment restrictions, is reshaping deal structures and contract negotiations. And data centers themselves are now delivered as integrated energy and infrastructure projects, with legal and commercial complexity to match.
New partnership structures are emerging
The sheer scale of the investment required to deliver next-generation data centers has made joint ventures a primary delivery model. Research estimates 28% annual compound growth in hyperscale investment globally over the next decade, with spending set to exceed USD1.5 trillion a year by 2034. In 2025 alone, Nvidia announced a USD100 billion investment program, while other U.S. technology companies together committed hundreds of billions more.
Three joint venture (JV) structures have emerged amid this expansion. The first pairs developers with financial sponsors (infrastructure funds, pension plans, or sovereign wealth investors) in programmatic or platform-level partnerships, exemplified by the USD7bn deal between Blackstone and Digital Realty.
The second combines developers and power counterparties, from independent power producers to equipment manufacturers, in line with the fact that securing megawatts is now as commercially decisive as securing land.
The third brings hyperscalers directly into the equity structure of digital infrastructure, with Meta’s USD27bn Hyperion venture with Blue Owl Capital a prime example. These models are not mutually exclusive; larger transactions increasingly combine elements of all three, assembling capital, power, and offtake in a single program.
Power as a binding constraint
Power scarcity is one of the critical headwinds facing the sector. Grid-connection queues can now stretch to ten years in parts of Europe, and U.S. interconnection upgrades persistently lag demand. By the end of the decade, analysts predict that U.S. data centers will consume more electricity than the country uses producing aluminum, steel, cement, chemicals and all other energy-intensive goods combined.
The IEA predicts that USD3.9tn will be invested in data centers globally between 2026 and 2030, yet the grid infrastructure required to deliver the power they need is in urgent need of expansion. Here, reinforcement programs carry their own multi-year lead times.
Behind-the-meter (BTM) generation has emerged as a route to faster and more reliable supply, but it brings its own financing, regulatory, and technology risks. A dedicated power plant co-located with the data center can sidestep congested grid queues and reduce exposure to volatile wholesale energy prices, yet the developer must reconcile the intermittency of renewables or the emissions profile of gas-fired generation with the “five-nines” uptime that data centers require (i.e., 99.999% availability).
Small modular nuclear reactors are increasingly being considered as a medium-term solution, though technology maturity and regulatory complexity present significant hurdles. Legacy regulatory frameworks designed for centralized grids have not been updated to accommodate private generation of the scale data centers need, creating licensing uncertainties that require jurisdiction-specific analysis.
Financing structures in transition
The way data centers are financed is undergoing its own transformation. Lenders have traditionally underwritten data centers as real estate, basing credit on the value of land and physical property. That model is giving way to a blended real estate and infrastructure approach that mirrors the changing risk profile of data center assets over their lifecycle, from acquisition through construction to long-term capital markets takeout.
The rise of AI-focused data centers, which require developers to own both shell and core (including graphics processing units (GPUs) and related hardware), adds further complexity. The core assets can be worth nearly four times the shell, introduce equipment obsolescence risk, and demand separate or layered financing structures.
Once operational, data centers with predictable lease revenues can access broadly syndicated term loans, private placements, and asset-backed securitizations, but hurdles remain. Mismatches between financing tenors and typical hyperscaler contract durations of around ten years introduce renewal risk.
Rating agency requirements for insolvency-remote structures can conflict with national security regimes. These dynamics are mirrored in the GPUaaS market, where the capital intensity of GPU procurement, combined with rapid hardware obsolescence, has attracted private credit funds, institutional lenders and specialist boutique financiers offering buy-and-lease structures. Across the sector, the traditional lines between real estate loans, infrastructure financing, structured credit and capital markets solutions are blurring, and a coherent capital strategy now requires careful planning across all of them.
Geopolitics, regulation, and cybersecurity
Data centers sit at the intersection of technology, national security and trade policy. The U.S. AI Action Plan, published in July 2025, seeks to remove regulatory barriers while exporting “full stack” AI packages to allied nations and restricting China’s access to advanced semiconductors.
The U.S. and allied nations are pursuing legal reforms that would extend export controls on GPUs, which have traditionally been focused on hardware shipments, to cloud-based compute. For GPUaaS providers and their customers, this evolving landscape requires robust contractual risk allocation, enhanced end-user due diligence and change-in-law protections.
Data sovereignty has become equally consequential. More than 100 countries have introduced data localization frameworks, and the EU’s Cloud and AI Development Act introduces a sovereignty framework and mandatory sovereignty risk assessments for public entities.
In response, hyperscalers are localizing infrastructure. Microsoft, for example, announced it would increase its European data center capacity by 40% by 2027 and store data within member states’ borders, with operators also building smaller, nationally oriented facilities rather than pan-Continental campuses.
The classification of data centers as critical infrastructure in the EU, the UK and across Asia Pacific brings stringent cybersecurity and resilience obligations that cannot be downstreamed to tenants. The convergence of IT and operational technology in modern facilities makes data centers uniquely exposed to sophisticated physical and cyberattacks; even brief disruptions can trigger missed service-level agreements and physical damage. For operators, disciplined execution of cyber preparedness measures, incident response planning and a clear understanding of legal and regulatory reporting obligations are now operational imperatives.
Environmental and social pressures
The sector's growing energy footprint has drawn concern from investors, local communities and regulators. In some jurisdictions, calls for moratoriums on data center expansion are gaining traction, fueled by the tension between the electricity requirements of AI-driven facilities and commitments to transition away from fossil fuels.
The EU’s Energy Efficiency Directive already requires data centers above 500 kW to report on energy and water efficiency, renewable energy use and waste heat recovery, and a sustainability rating scheme is expected to be adopted in the second half of 2026.
Tighter environmental and human rights impact assessments, conditional permitting and rising engagement from supply chain actors and financiers are likely to intensify. Market participants negotiating joint ventures, leases, and financing arrangements will need to anticipate these pressures through contractual provisions that allocate sustainability compliance obligations and reporting requirements with precision.
The common thread that links all these dynamics is complexity: securing power means navigating grid queues and BTM regulatory gaps; deploying capital means blending real estate, infrastructure, and equipment financing in novel structures; and operating across borders means contending with divergent export controls, data sovereignty regimes, and sustainability mandates.
For data center developers, the strategic question is no longer whether to take a partner, but who to collaborate with given the constraints of a particular site, jurisdiction, and customer base.
For investors, lenders and hyperscalers, the discipline of allocating risk across multiple counterparties, as well as an increasingly challenging regulatory environment, will define which projects move from announcement to operation.
