Asia Pacific is entering a data-center investment cycle that could require $772 billion of capital through 2030, according to JLL’s high-growth scenario. The forecast calls for 24 gigawatts of additional capacity between 2025 and 2030, a scale equivalent to about 240 100-megawatt campuses. Of that investment, roughly $286 billion would flow into data-center real estate, while another $486 billion would support GPUs, servers, networking systems and related computing infrastructure. The numbers capture more than an expansion of digital real estate because they point to a fundamental shift in how investors must think about compute assets. Asia’s challenge is no longer simply finding customers for new facilities; it is converting enormous pools of capital into powered, connected and technologically relevant capacity before demand moves elsewhere.

The opportunity sits within a broader infrastructure cycle accelerated by artificial intelligence, cloud adoption and rising digital consumption across the region. PwC expects Asia Pacific to account for $8.2 trillion in cumulative data-center capital expenditure through 2050, with China and India representing major sources of that demand. Globally, the firm estimates cumulative spending could reach $31.6 trillion under its central scenario, with an upside case approaching $50 trillion if AI adoption advances faster than expected. That outlook changes the investment equation because computing equipment requires repeated reinvestment even when the underlying building remains productive. PwC expects ICT equipment to rise from about 70% of data-center investment today to 93% by 2050, reflecting replacement cycles for servers, GPUs and networking equipment that typically run only four to six years.

Power Becomes the Core Constraint

Demand for capacity appears considerably stronger than the supply pipeline might suggest. JLL expects Asia Pacific to add 4.8 gigawatts of new data-center supply by 2027, with about 78% already preleased. Regional vacancy could remain around 6.5% to 7% for several years as hyperscalers absorb much of the incoming capacity. CBRE estimates that technology companies will increase AI-related capital expenditure by 61% in 2026, while average new data-center developments across Asia Pacific have grown beyond 100 megawatts. Those figures support the investment case, but they do not resolve the physical bottleneck confronting developers. A signed lease can establish revenue visibility, yet it cannot by itself secure electricity, accelerate permitting or guarantee that a project will reach commercial operation on schedule.

Power has consequently become a strategic asset rather than a development input. JLL estimates that obtaining a grid connection can take about 24 months in some emerging markets and more than eight years in certain established locations. The disparity creates a structural problem because grid-connection and transmission timelines can materially constrain the pace at which new data-center capacity reaches operation. A project that reaches mechanical completion without dependable power remains unable to monetize its most valuable resource: computing capacity. As a result, sites with secured and deliverable electricity are commanding greater strategic importance than land parcels supported only by future grid availability.

Southeast Asia Faces a New Energy Test

Southeast Asia illustrates the tension between strong digital demand and constrained infrastructure. Data-center electricity consumption across Asia Pacific almost doubled from 2020 to 2024 and is expected to triple over the next several years, according to CBRE, with power availability emerging as a major constraint on regional expansion. CBRE estimates that data-center electricity consumption across Asia Pacific nearly doubled from 2020 to 2024 and could triple over the coming years. That acceleration reflects cloud expansion, AI workloads and increasing deployment density rather than conventional enterprise demand alone. Higher-density infrastructure will place new requirements on electricity networks, cooling systems and water resources.

Malaysia is taking a different approach by expanding the available regional footprint. The Malaysian Investment Development Authority approved RM144.4 billion, or about $35.7 billion, in data-center and cloud-computing investment from 2021 through mid-2025. Data centers and cloud infrastructure accounted for 76.8% of approved digital investment in 2024, according to the Malaysia Digital Economy Corporation. Johor has consequently emerged as an important extension of the Singapore ecosystem, combining greater land availability with a larger potential power base. Meanwhile, the cross-border relationship creates a distinctive investment model in which Singapore supplies connectivity, customers and financial depth while neighboring markets absorb some of the physical expansion.

Asia’s Markets Are Not Interchangeable

The regional opportunity extends well beyond Singapore and Malaysia. India is attracting major cloud and AI commitments on the back of its large domestic digital economy and government-backed technology ambitions. Japan and Australia offer deeper institutional markets and established infrastructure ecosystems, although their development constraints differ from those of emerging Southeast Asian hubs. Indonesia, Thailand, the Philippines and Vietnam are also building positions in the regional compute economy, but each market presents a different combination of energy availability, connectivity, regulation and customer demand. Investors therefore cannot treat Asia Pacific as one homogeneous data-center market simply because capital and technology companies operate across national borders.

Operating assets in established markets continue to attract substantial institutional interest. Keppel DC REIT and Keppel agreed this week to acquire two Tokyo data centers for $1.19 billion, demonstrating continued appetite for stabilized digital infrastructure in mature locations. Yet investors buying operational facilities face a fundamentally different risk profile from developers assembling large campuses in emerging markets. Construction capability, foreign-ownership rules, currency exposure, permitting regimes and hyperscale customer concentration can materially change expected returns. The same headline megawatt capacity can therefore represent very different economics depending on where it sits and how quickly it can become revenue-generating infrastructure.

AI Changes the Data-Center Capital Stack

The $772 billion estimate contains a particularly important distinction: much of the required investment will go into computing equipment rather than property. JLL’s projections indicate that up to $486 billion of the $772 billion capital requirement could be needed to fit out data centers with GPUs and networking infrastructure. That allocation creates a mismatch between asset lives because a data-center building and its electrical systems can operate for decades, while GPUs and servers may require replacement within four to six years. The financing structure must therefore account for two different depreciation and technology cycles inside one broader infrastructure platform. Traditional real-estate underwriting alone cannot capture that complexity.

Neoclouds and specialized GPU platforms are emerging in response to the shortage of high-performance computing outside the largest hyperscalers. These companies are emerging as a distinct demand category for high-performance computing, while their rapid expansion creates new questions around customer concentration, utilization and the pace of hardware deployment. They must assess customer concentration, utilization, contract duration, hardware resale values and the pace of technological obsolescence. A GPU platform can generate strong demand today while still carrying significant refinancing risk if its equipment becomes outdated before the underlying customer contracts expire.

Construction Capital Is Getting More Complex

Data-center development itself is becoming more capital intensive. Large AI campuses require significant expenditure before they begin producing cash flow, while land, electrical equipment, advanced cooling systems and sustainability requirements continue to increase costs. Construction schedules also face competition for specialist components and skilled contractors. Developers that once relied primarily on equity and bank facilities are now looking across private credit, institutional capital, asset-backed structures, credit enhancements and public debt markets.

The financing strategy may also need to evolve after construction. Bridge or development capital can support a campus through the construction period, but stabilized facilities may ultimately require cheaper long-duration financing. That transition creates opportunities for institutional investors, private lenders and public markets while placing greater emphasis on contract quality and operational performance. A hyperscale-leased facility with predictable cash flow should command a different cost of capital from a campus awaiting its grid connection. Likewise, a GPU platform built around rapidly changing computing equipment should not automatically receive the same valuation framework as a mature colocation asset.

These distinctions will become increasingly important as developers move from individual projects toward multi-gigawatt platforms. Larger portfolios can diversify customer and construction risk, but they also require significantly more capital before the platform reaches scale. The ability to coordinate financing across land, power infrastructure, buildings and computing equipment could become a competitive advantage in its own right. Capital providers that understand those interdependencies will have more influence over which projects reach completion.

Governments Are Becoming Infrastructure Investors

The data-center buildout is also becoming part of national technology strategy. Governments increasingly regard computing capacity, semiconductor access and domestic infrastructure as strategic assets tied to economic competitiveness and resilience. That perspective is encouraging sovereign AI initiatives and bringing public institutions deeper into decisions over land, power allocation, data residency and infrastructure ownership. Government participation can accelerate development, but it can also introduce requirements that change project economics.

Trade policy could further reshape the capital map. PwC estimates that tighter export controls disrupting global chip supply chains could reduce cumulative global AI infrastructure investment to about $25.5 trillion through 2050, roughly $6 trillion below its $31.6 trillion central forecast. Countries seeking greater control over their computing capacity may favor domestic infrastructure, local partnerships or sovereign-backed projects even when those structures carry higher costs. Investors must therefore consider semiconductor availability and national technology policy alongside conventional infrastructure fundamentals.

Indonesia Investment Authority CEO Oki Ramadhana is scheduled to present a sovereign investor’s perspective on digital infrastructure at The Tech Capital’s APAC Finance Forum. Singapore Senior Minister of State for Digital Development and Information Tan Kiat How is also among the confirmed participants. Their involvement reflects a market in which government, capital and technology policy increasingly intersect. Ultimately, the question is not simply whether Asia can attract enough money, but whether policymakers and investors can coordinate that money with the physical systems required to turn it into dependable compute.

The Investment Test Is Execution

The region’s projected investment requirement should not be interpreted as a guarantee that every proposed project will proceed. Large pipeline numbers can conceal delays caused by power shortages, permitting, equipment constraints, financing gaps or shifts in technology demand. The more important investment question is which projects can demonstrate a credible path from land acquisition to powered operations and contracted revenue. That distinction could determine where returns accumulate as Asia’s compute market matures.

The forum will also address GPU financing, credit wraps, development capital, private lending, Rule 144A markets and potential IPO routes, with Davis Polk hosting a private executive discussion. Another plenary will examine how data centers, power generation, connectivity and computing equipment can fit into an emerging capital stack. Executive Pass holders will have access to private discussions covering energy financing, hyperscale AI campus delivery, and GPU and data-center infrastructure funding under Chatham House Rule. The breadth of those discussions reflects a market that increasingly requires investors to understand technology cycles and energy systems alongside conventional infrastructure finance.

Asia’s $772 billion data-center requirement ultimately represents an execution challenge as much as a capital opportunity. Demand for AI and cloud capacity can justify enormous investment, but only reliable power, efficient construction, strong connectivity and durable financing can turn that demand into productive assets. The region now has the prospect of becoming one of the world’s largest compute markets, but its ability to capture that opportunity will depend on how effectively it resolves the infrastructure bottlenecks standing between capital commitments and live capacity.