Technology
Data Centres: How AI Infrastructure Is Creating Real Estate's Most Compelling Investment Case
By Abhii Dabas
April 29, 2026
10 min read
Introduction
Few real estate sectors have ever attracted capital at the velocity now directed at data centres. The AI infrastructure buildout — led by Alphabet, Amazon, Microsoft, and Meta collectively investing approximately $400 billion in data centre capacity in 2026 alone — is creating a demand wave that is reshaping entire electricity grids, redirecting institutional capital flows, and generating a new class of real estate asset with financial characteristics unlike anything the traditional property industry has encountered. Total anticipated investment in the sector through 2030 stands at $3 trillion, encompassing $1.2 trillion in real estate asset value creation, $870 billion in new debt financing, and $1–2 trillion in tenant IT equipment. For real estate investors, data centres offer stabilised net operating income yields exceeding 10%, development margins north of 50%, and long-duration lease structures with investment-grade hyperscaler covenants — a combination that has compressed implied cap rates to 4.4% for listed vehicle exposure and generated intense competition for ground-up development opportunities wherever power connectivity exists.
The AI Demand Inflection: Why This Cycle Is Different
- From Cloud to AI — A Step-Change in Compute Intensity:The cloud computing buildout of 2015–2022 drove the first wave of data centre institutionalisation. The AI infrastructure buildout of 2023–2030 is categorically more capital-intensive. AI training workloads require GPU clusters with power densities 5–10 times higher than conventional cloud servers; a single AI training facility can consume 100–300 megawatts of power, equivalent to the residential demand of a mid-sized city. Currently, AI workloads account for approximately 25% of all data centre compute; by 2030, JLL projects this figure to reach 50%. Goldman Sachs Research estimates total data centre demand will grow by approximately 50% to 92 gigawatts by 2027 alone, with AI as the primary driver of incremental demand beyond baseline cloud growth.
- Hyperscaler Capex — The Numbers Behind the Demand:The four major hyperscalers — Alphabet, Amazon, Microsoft, and Meta — collectively invested over $350 billion in data centre infrastructure during 2025 and have committed approximately $400 billion for 2026. This is not a uniform global footprint: North American capacity is being augmented by rapid expansion into Northern Virginia, the Phoenix cluster, Texas, and the Pacific Northwest, while EMEA investments concentrate on Dublin, Amsterdam, Frankfurt, London, and emerging markets including Warsaw, Madrid, and the Nordics. Microsoft's $80 billion+ annual data centre capex programme has driven acquisition of land banks in markets previously overlooked by institutional investors, fundamentally repricing land with power connectivity.
- From Training to Inference — The Next Regional Demand Driver:The data centre investment narrative of 2023–2025 was dominated by AI training — large-scale GPU clusters in a handful of hyperscale campus markets. By 2027, JLL projects inference workloads will overtake training as the primary driver of new capacity. This shift is strategically significant for real estate investors: inference infrastructure requires low-latency proximity to end users, driving demand into secondary and regional markets that training campuses bypassed entirely. Cities including Singapore, Sydney, Tokyo, São Paulo, Mumbai, and Riyadh are seeing accelerating data centre development specifically to serve AI inference delivery to local populations — a structural shift that creates investable opportunities far beyond the established Tier 1 markets.
Market Scale: $3 Trillion and What It Means for Real Estate
- Capacity Growth — 100GW in Five Years:Global data centre capacity is projected to reach 200 gigawatts by 2030, nearly doubling from current installed base. Nearly 100GW of new capacity will be added between 2026 and 2030 — a construction programme requiring real estate assets of unprecedented scale and specification. The sector is growing at a 14% CAGR through 2030, with the Americas leading at 17% CAGR (driven by US hyperscaler demand), Asia Pacific at 12% (32GW to 57GW by 2030), and EMEA at 10% CAGR with a particular focus on sovereign AI cloud infrastructure for European governments. Construction costs are rising at 7% annually, from $10.7 million to $11.3 million per megawatt, reflecting labour shortages, specialised equipment lead times, and rising structural steel costs.
- Real Estate Asset Value Creation:JLL estimates that delivering 100GW of new capacity through 2030 will create $1.2 trillion in new real estate asset value — the largest single sector-driven asset creation in commercial real estate history over a comparable timeframe. This value creation is not evenly distributed: land parcels with existing grid connection or behind-the-meter power generation that would previously trade as industrial or logistics assets are now commanding data centre land premiums of 3–10x their prior use valuation. Markets with renewable energy abundance (Nordics, Pacific Northwest, Iceland) are particularly favoured as operators seek to meet corporate sustainability commitments while securing low-cost, long-term power.
- Investment Returns — Superior Yields With Complexity Premium:Stabilised data centre net operating income yields exceeding 10% compare favourably against the 5–7% yields typical of premium logistics and core office assets. Development margins above 50% reflect both the specialised construction complexity and the locked-in lease structures that hyperscalers execute before ground break. Listed data centre REITs — led by Equinix and Digital Realty — have compressed to implied cap rates of 4.4%, with Equinix projecting 9–10% revenue growth in 2026 and AFFO per share growth of 8–10%, and Digital Realty targeting core FFO per share of $7.90–$8.00 in 2026 (approximately 8% growth). The premium to direct-ownership yields reflects REITs' liquidity, diversification, and platform scale advantages.
Power: The Binding Constraint on Capital Deployment
- The Grid Queue — Four-Year Waits in Primary Markets:The single largest constraint on data centre development in 2026 is not capital, construction labour, or tenant demand — it is grid connectivity. In primary data centre markets including Northern Virginia, Frankfurt, Dublin, Amsterdam, and Singapore, connection requests to national grids now face wait times exceeding four years. This constraint has effectively bifurcated the market: operators with existing grid reservations hold a structural advantage that is increasingly difficult for new entrants to replicate. Developers who secured grid capacity in 2020–2022 — when demand seemed speculative — are now sitting on optionality worth billions. For investors entering today, power security must be verified as the first item of due diligence, not an afterthought.
- Behind-the-Meter Solutions and the Energy Transition:Operators unable to access grid power at required scale are increasingly deploying behind-the-meter generation, including large-scale natural gas turbines in the Americas, battery storage combined with renewable power purchase agreements in EMEA, and diesel backup systems in emerging markets awaiting grid upgrades. Some markets have introduced "bring your own power" mandates — requiring data centre developers to demonstrate independent power generation capacity before planning approval — fundamentally altering the development economics and site selection criteria. Nuclear power proximity is emerging as a data centre site selection advantage: Microsoft's reactivation deal with Three Mile Island and Amazon's nuclear site purchases signal that hyperscalers view large-scale nuclear as the long-term answer to reliable, low-carbon baseload power.
- Water and Cooling Innovation:Data centres' water consumption for cooling has emerged as a material planning and ESG concern in markets experiencing increasing water stress. A large AI training cluster can consume millions of litres of water daily for cooling towers. Liquid cooling technology — both direct-to-chip and immersion cooling — is reducing air and water cooling demands by 30–40% while enabling higher power densities, and is rapidly becoming the specification standard for new AI-optimised facilities. Investors should differentiate between legacy air-cooled data centres (increasingly challenged to accommodate AI tenant requirements) and liquid-cooled next-generation facilities built to the compute density standards AI workloads require.
Regional Opportunities and Key Markets
- Northern Virginia — Hyperscale Capital of the World:Loudoun County, Virginia remains the world's largest data centre concentration, hosting over 25% of global hyperscale capacity. Land values in the Ashburn corridor have appreciated 400–600% over the past decade, with remaining developable land commanding $2–4 million per acre against sub-$200,000 five years ago. The market continues to attract investment despite maturity: grid expansion programmes by Dominion Energy are unlocking new development areas, and the county's fibre connectivity, tax incentives, and established operator ecosystem create network effects that are difficult to replicate elsewhere. For institutional investors, stabilised Northern Virginia assets offer the sector's most liquid exit market.
- EMEA — Sovereign AI Cloud and Emerging Markets:Europe's data centre growth is being shaped by two concurrent forces: hyperscaler demand concentrated in the established FLAP markets (Frankfurt, London, Amsterdam, Paris) and a new wave of sovereign AI cloud investment from European governments seeking data sovereignty and AI capability outside US-headquartered infrastructure. The latter is creating greenfield development opportunities in markets including Warsaw, Madrid, Vienna, Stockholm, and Helsinki, where government anchor tenants provide underwriting certainty for developer risk. EMEA adds approximately 13GW of new capacity through 2030, with renewable energy abundance in the Nordics attracting carbon-sensitive hyperscaler tenants willing to pay site development premiums for clean power certification.
- Asia Pacific — Inference Demand Driving Tier 2 Markets:Asia Pacific data centre capacity grows from 32GW to 57GW by 2030 (12% CAGR), with Japan, Australia, India, and Southeast Asia all recording accelerating development pipelines. Singapore's government data centre moratorium — introduced to manage power and water consumption — has redirected regional hyperscaler investment to Johor Bahru (Malaysia), Jakarta (Indonesia), and Bangkok (Thailand), creating new institutional investment markets where land values and development margins remain attractive relative to Singapore's constrained supply. India's data centre sector is experiencing a particularly significant inflection, with hyperscaler investment aligned with government digital public infrastructure ambitions creating a pipeline JLL estimates will surpass Singapore's total installed base within three years.
Investment Framework: Accessing the Data Centre Opportunity
- Entry Strategies — Listed, Unlisted, and Direct:Data centre investment is accessible across the liquidity spectrum. Listed REITs (Equinix, Digital Realty) provide liquid, diversified exposure at compressed cap rates — appropriate as a liquid anchor position. Unlisted core infrastructure funds targeting stabilised data centres offer yields of 6–8% with less liquidity but greater return stability. Direct development — either as an operator/developer or through a joint venture with an established platform — offers the highest potential returns (development margins >50%) but requires deep technical expertise, power procurement capability, and tenant relationships. Most institutional investors entering the sector for the first time should partner with an established platform operator rather than attempt independent development.
- Underwriting Discipline in a Hot Market:The combination of extraordinary demand, compressed cap rates, and intense institutional competition has created conditions where disciplined underwriting can be overlooked. Key risks to stress-test include: power procurement timeline and cost certainty; tenant credit quality and lease length (hyperscaler 10–20 year leases are fundamentally different from enterprise 3–5 year deals); technology obsolescence risk (a facility built for 2025 AI density requirements may require significant capital expenditure to accommodate 2030 density standards); and exit liquidity, which in secondary markets may be significantly thinner than assumptions imply. The sector's complexity premium is real — and so is the execution risk for investors without platform capabilities.
This content is AI-generated and may contain errors. Figures are indicative and subject to change. Do your own due diligence and seek independent legal and financial advice.
Sources
- JLL 2026 Global Data Center Outlook
- Data Centre Knowledge — CRE's Jewel for 2026
- Data Centre Magazine — 200GW Capacity by 2030
- SitusAMC — Data Centers in an AI-Driven Era: Key Trends Reshaping Valuation
- CRE Daily — Data Center Investment Surges Amid AI Boom
- Motley Fool — Best Data Center REITs for 2026
- Avid Solutions — 13 Data Center Growth Projections 2026-2030
Author
Abhii DabasFounder & CEO, INTRIC Global
Abhii Dabas is the Founder and CEO of INTRIC Global, the cross-border property intelligence platform for serious investors. He advises high-net-worth buyers on international real estate strategy and has evaluated residential markets across more than 40 countries.
Share
Share this article with others
Introduction
Few real estate sectors have ever attracted capital at the velocity now directed at data centres. The AI infrastructure buildout — led by Alphabet, Amazon, Microsoft, and Meta collectively investing approximately $400 billion in data centre capacity in 2026 alone — is creating a demand wave that is reshaping entire electricity grids, redirecting institutional capital flows, and generating a new class of real estate asset with financial characteristics unlike anything the traditional property industry has encountered. Total anticipated investment in the sector through 2030 stands at $3 trillion, encompassing $1.2 trillion in real estate asset value creation, $870 billion in new debt financing, and $1–2 trillion in tenant IT equipment. For real estate investors, data centres offer stabilised net operating income yields exceeding 10%, development margins north of 50%, and long-duration lease structures with investment-grade hyperscaler covenants — a combination that has compressed implied cap rates to 4.4% for listed vehicle exposure and generated intense competition for ground-up development opportunities wherever power connectivity exists.
The AI Demand Inflection: Why This Cycle Is Different
- From Cloud to AI — A Step-Change in Compute Intensity:The cloud computing buildout of 2015–2022 drove the first wave of data centre institutionalisation. The AI infrastructure buildout of 2023–2030 is categorically more capital-intensive. AI training workloads require GPU clusters with power densities 5–10 times higher than conventional cloud servers; a single AI training facility can consume 100–300 megawatts of power, equivalent to the residential demand of a mid-sized city. Currently, AI workloads account for approximately 25% of all data centre compute; by 2030, JLL projects this figure to reach 50%. Goldman Sachs Research estimates total data centre demand will grow by approximately 50% to 92 gigawatts by 2027 alone, with AI as the primary driver of incremental demand beyond baseline cloud growth.
- Hyperscaler Capex — The Numbers Behind the Demand:The four major hyperscalers — Alphabet, Amazon, Microsoft, and Meta — collectively invested over $350 billion in data centre infrastructure during 2025 and have committed approximately $400 billion for 2026. This is not a uniform global footprint: North American capacity is being augmented by rapid expansion into Northern Virginia, the Phoenix cluster, Texas, and the Pacific Northwest, while EMEA investments concentrate on Dublin, Amsterdam, Frankfurt, London, and emerging markets including Warsaw, Madrid, and the Nordics. Microsoft's $80 billion+ annual data centre capex programme has driven acquisition of land banks in markets previously overlooked by institutional investors, fundamentally repricing land with power connectivity.
- From Training to Inference — The Next Regional Demand Driver:The data centre investment narrative of 2023–2025 was dominated by AI training — large-scale GPU clusters in a handful of hyperscale campus markets. By 2027, JLL projects inference workloads will overtake training as the primary driver of new capacity. This shift is strategically significant for real estate investors: inference infrastructure requires low-latency proximity to end users, driving demand into secondary and regional markets that training campuses bypassed entirely. Cities including Singapore, Sydney, Tokyo, São Paulo, Mumbai, and Riyadh are seeing accelerating data centre development specifically to serve AI inference delivery to local populations — a structural shift that creates investable opportunities far beyond the established Tier 1 markets.
Market Scale: $3 Trillion and What It Means for Real Estate
- Capacity Growth — 100GW in Five Years:Global data centre capacity is projected to reach 200 gigawatts by 2030, nearly doubling from current installed base. Nearly 100GW of new capacity will be added between 2026 and 2030 — a construction programme requiring real estate assets of unprecedented scale and specification. The sector is growing at a 14% CAGR through 2030, with the Americas leading at 17% CAGR (driven by US hyperscaler demand), Asia Pacific at 12% (32GW to 57GW by 2030), and EMEA at 10% CAGR with a particular focus on sovereign AI cloud infrastructure for European governments. Construction costs are rising at 7% annually, from $10.7 million to $11.3 million per megawatt, reflecting labour shortages, specialised equipment lead times, and rising structural steel costs.
- Real Estate Asset Value Creation:JLL estimates that delivering 100GW of new capacity through 2030 will create $1.2 trillion in new real estate asset value — the largest single sector-driven asset creation in commercial real estate history over a comparable timeframe. This value creation is not evenly distributed: land parcels with existing grid connection or behind-the-meter power generation that would previously trade as industrial or logistics assets are now commanding data centre land premiums of 3–10x their prior use valuation. Markets with renewable energy abundance (Nordics, Pacific Northwest, Iceland) are particularly favoured as operators seek to meet corporate sustainability commitments while securing low-cost, long-term power.
- Investment Returns — Superior Yields With Complexity Premium:Stabilised data centre net operating income yields exceeding 10% compare favourably against the 5–7% yields typical of premium logistics and core office assets. Development margins above 50% reflect both the specialised construction complexity and the locked-in lease structures that hyperscalers execute before ground break. Listed data centre REITs — led by Equinix and Digital Realty — have compressed to implied cap rates of 4.4%, with Equinix projecting 9–10% revenue growth in 2026 and AFFO per share growth of 8–10%, and Digital Realty targeting core FFO per share of $7.90–$8.00 in 2026 (approximately 8% growth). The premium to direct-ownership yields reflects REITs' liquidity, diversification, and platform scale advantages.
Power: The Binding Constraint on Capital Deployment
- The Grid Queue — Four-Year Waits in Primary Markets:The single largest constraint on data centre development in 2026 is not capital, construction labour, or tenant demand — it is grid connectivity. In primary data centre markets including Northern Virginia, Frankfurt, Dublin, Amsterdam, and Singapore, connection requests to national grids now face wait times exceeding four years. This constraint has effectively bifurcated the market: operators with existing grid reservations hold a structural advantage that is increasingly difficult for new entrants to replicate. Developers who secured grid capacity in 2020–2022 — when demand seemed speculative — are now sitting on optionality worth billions. For investors entering today, power security must be verified as the first item of due diligence, not an afterthought.
- Behind-the-Meter Solutions and the Energy Transition:Operators unable to access grid power at required scale are increasingly deploying behind-the-meter generation, including large-scale natural gas turbines in the Americas, battery storage combined with renewable power purchase agreements in EMEA, and diesel backup systems in emerging markets awaiting grid upgrades. Some markets have introduced "bring your own power" mandates — requiring data centre developers to demonstrate independent power generation capacity before planning approval — fundamentally altering the development economics and site selection criteria. Nuclear power proximity is emerging as a data centre site selection advantage: Microsoft's reactivation deal with Three Mile Island and Amazon's nuclear site purchases signal that hyperscalers view large-scale nuclear as the long-term answer to reliable, low-carbon baseload power.
- Water and Cooling Innovation:Data centres' water consumption for cooling has emerged as a material planning and ESG concern in markets experiencing increasing water stress. A large AI training cluster can consume millions of litres of water daily for cooling towers. Liquid cooling technology — both direct-to-chip and immersion cooling — is reducing air and water cooling demands by 30–40% while enabling higher power densities, and is rapidly becoming the specification standard for new AI-optimised facilities. Investors should differentiate between legacy air-cooled data centres (increasingly challenged to accommodate AI tenant requirements) and liquid-cooled next-generation facilities built to the compute density standards AI workloads require.
Regional Opportunities and Key Markets
- Northern Virginia — Hyperscale Capital of the World:Loudoun County, Virginia remains the world's largest data centre concentration, hosting over 25% of global hyperscale capacity. Land values in the Ashburn corridor have appreciated 400–600% over the past decade, with remaining developable land commanding $2–4 million per acre against sub-$200,000 five years ago. The market continues to attract investment despite maturity: grid expansion programmes by Dominion Energy are unlocking new development areas, and the county's fibre connectivity, tax incentives, and established operator ecosystem create network effects that are difficult to replicate elsewhere. For institutional investors, stabilised Northern Virginia assets offer the sector's most liquid exit market.
- EMEA — Sovereign AI Cloud and Emerging Markets:Europe's data centre growth is being shaped by two concurrent forces: hyperscaler demand concentrated in the established FLAP markets (Frankfurt, London, Amsterdam, Paris) and a new wave of sovereign AI cloud investment from European governments seeking data sovereignty and AI capability outside US-headquartered infrastructure. The latter is creating greenfield development opportunities in markets including Warsaw, Madrid, Vienna, Stockholm, and Helsinki, where government anchor tenants provide underwriting certainty for developer risk. EMEA adds approximately 13GW of new capacity through 2030, with renewable energy abundance in the Nordics attracting carbon-sensitive hyperscaler tenants willing to pay site development premiums for clean power certification.
- Asia Pacific — Inference Demand Driving Tier 2 Markets:Asia Pacific data centre capacity grows from 32GW to 57GW by 2030 (12% CAGR), with Japan, Australia, India, and Southeast Asia all recording accelerating development pipelines. Singapore's government data centre moratorium — introduced to manage power and water consumption — has redirected regional hyperscaler investment to Johor Bahru (Malaysia), Jakarta (Indonesia), and Bangkok (Thailand), creating new institutional investment markets where land values and development margins remain attractive relative to Singapore's constrained supply. India's data centre sector is experiencing a particularly significant inflection, with hyperscaler investment aligned with government digital public infrastructure ambitions creating a pipeline JLL estimates will surpass Singapore's total installed base within three years.
Investment Framework: Accessing the Data Centre Opportunity
- Entry Strategies — Listed, Unlisted, and Direct:Data centre investment is accessible across the liquidity spectrum. Listed REITs (Equinix, Digital Realty) provide liquid, diversified exposure at compressed cap rates — appropriate as a liquid anchor position. Unlisted core infrastructure funds targeting stabilised data centres offer yields of 6–8% with less liquidity but greater return stability. Direct development — either as an operator/developer or through a joint venture with an established platform — offers the highest potential returns (development margins >50%) but requires deep technical expertise, power procurement capability, and tenant relationships. Most institutional investors entering the sector for the first time should partner with an established platform operator rather than attempt independent development.
- Underwriting Discipline in a Hot Market:The combination of extraordinary demand, compressed cap rates, and intense institutional competition has created conditions where disciplined underwriting can be overlooked. Key risks to stress-test include: power procurement timeline and cost certainty; tenant credit quality and lease length (hyperscaler 10–20 year leases are fundamentally different from enterprise 3–5 year deals); technology obsolescence risk (a facility built for 2025 AI density requirements may require significant capital expenditure to accommodate 2030 density standards); and exit liquidity, which in secondary markets may be significantly thinner than assumptions imply. The sector's complexity premium is real — and so is the execution risk for investors without platform capabilities.
This content is AI-generated and may contain errors. Figures are indicative and subject to change. Do your own due diligence and seek independent legal and financial advice.
Sources
- JLL 2026 Global Data Center Outlook
- Data Centre Knowledge — CRE's Jewel for 2026
- Data Centre Magazine — 200GW Capacity by 2030
- SitusAMC — Data Centers in an AI-Driven Era: Key Trends Reshaping Valuation
- CRE Daily — Data Center Investment Surges Amid AI Boom
- Motley Fool — Best Data Center REITs for 2026
- Avid Solutions — 13 Data Center Growth Projections 2026-2030
Author
Abhii DabasFounder & CEO, INTRIC Global
Abhii Dabas is the Founder and CEO of INTRIC Global, the cross-border property intelligence platform for serious investors. He advises high-net-worth buyers on international real estate strategy and has evaluated residential markets across more than 40 countries.
Share this article
Share this insight with others
Share
Share this article with others
Related Market Insights
View More Market Insights
TAX & RESIDENCY
Cross-Border Property Tax Guide for Singapore, UAE, and UK-Based Investors in 2026
By Abhii Dabas · July 27, 2026
Investment Guide
Why Gulf HNW Investors Still Buy in London: The Case for UK Property in 2026
By Abhii Dabas · July 21, 2026
Investment Guide
Buying Property in Australia in 2026: A Guide for Indonesian and Malaysian Investors
By Abhii Dabas · July 17, 2026
Found this useful? Send it to someone who should read it.
Share
Continue with INTRIC
