The AI revolution: How infrastructure is adapting
The AI revolution: How infrastructure is adapting
In a podcast as part of PEI Group's Private Markets 2030 programme, guests from Brookfield discussed the opportunities and risks surrounding artificial intelligence in infrastructure. The discussion highlighted the significant adaptations required in physical and digital infrastructure to support the rapid advancements and demands of AI technologies.
Context & What Changed
The advent and rapid acceleration of Artificial Intelligence (AI) technologies represent a fundamental shift in global technological and economic landscapes. Historically, technological revolutions have been underpinned by corresponding advancements in foundational infrastructure. The current AI revolution is no exception, but its demands are unprecedented in scale and complexity, particularly concerning computational power, data storage, and energy consumption. Traditional infrastructure, designed for previous generations of computing and data processing, is proving increasingly inadequate to meet the burgeoning requirements of AI models, which are characterized by their immense appetite for processing large datasets and executing complex algorithms (source: openai.com, deepmind.com). This necessitates a profound adaptation across various infrastructure sectors, including digital infrastructure (data centers, fiber optics), energy grids, and even physical real estate and water management. The core change is a shift from infrastructure primarily supporting human-driven data generation and consumption to one that must support machine-driven, high-intensity computational workloads. This requires not merely incremental upgrades but often entirely new paradigms in design, deployment, and operation, driven by the need for low-latency, high-bandwidth connectivity, and reliable, sustainable power at scale (source: infrastructureinvestor.com).
Stakeholders
The adaptation of infrastructure for AI involves a diverse ecosystem of stakeholders, each with distinct roles and interests:
Governments and Public Sector Agencies: Responsible for policy formulation, regulatory frameworks, land use planning, energy policy, environmental standards, and sometimes direct investment in critical infrastructure. Their role is pivotal in creating an enabling environment for AI infrastructure development, balancing economic growth with social and environmental considerations (source: ec.europa.eu, whitehouse.gov).
Infrastructure Developers and Investors (e.g., Brookfield): Private equity firms, pension funds, and sovereign wealth funds are increasingly allocating capital to digital infrastructure assets such as data centers, fiber networks, and renewable energy projects. They seek long-term, stable returns from essential services, driving the financing and construction of new facilities (source: brookfield.com, infrastructureinvestor.com).
Energy Companies (Utilities, Generators): Face immense pressure to provide reliable, high-capacity, and increasingly sustainable power to data centers. This includes traditional fossil fuel generators, renewable energy developers (solar, wind, hydro), and grid operators responsible for transmission and distribution (source: iea.org).
Data Center Operators: The direct consumers of vast amounts of energy and land, responsible for building, maintaining, and operating the physical facilities that house AI compute. They are at the forefront of innovation in cooling technologies, energy efficiency, and modular design (source: equinix.com, digitalrealty.com).
Telecommunications Providers: Essential for the high-speed, low-latency connectivity required to move data to and from AI models. This includes fiber optic network owners, 5G/6G wireless providers, and satellite communication companies (source: gsma.com, itu.int).
Technology Firms (AI Developers and Cloud Providers): Companies like Google, Microsoft, Amazon, Meta, and OpenAI are the primary drivers of AI innovation and demand. They build and operate hyperscale data centers or lease capacity, pushing the boundaries of computational infrastructure (source: google.com, microsoft.com, amazon.com).
Regulators: Agencies responsible for environmental protection, competition, data privacy, and energy markets. Their decisions significantly impact the feasibility, cost, and sustainability of AI infrastructure projects (source: epa.gov, ftc.gov).
Public Finance Institutions: Multilateral development banks, national development banks, and export credit agencies can play a role in de-risking investments, providing concessional financing, and supporting cross-border infrastructure projects (source: worldbank.org, eib.org).
Evidence & Data
The escalating demands of AI are evidenced by several key trends:
Data Center Growth: The global data center market size was valued at approximately USD 240 billion in 2023 and is projected to grow significantly, driven by AI and cloud computing (source: grandviewresearch.com, author's assumption for specific figure, but trend is verifiable). Hyperscale data centers, specifically designed for large-scale AI workloads, are proliferating globally, requiring substantial upfront capital investment (source: synergyresearchgroup.com).
Energy Consumption: AI training and inference are incredibly energy-intensive. Data centers already account for an estimated 1-1.5% of global electricity demand, and this share is projected to rise sharply with AI adoption (source: iea.org). Some estimates suggest that by 2030, AI could consume as much electricity as entire countries (source: nature.com, author's assumption for specific projection, but trend is verifiable). This necessitates massive investments in new power generation, particularly renewables, and grid upgrades (source: iea.org).
Water Usage: Advanced cooling systems for data centers, especially those housing high-density AI hardware, consume significant amounts of water. This creates pressure on local water resources, particularly in drought-prone regions (source: google.com/sustainability, microsoft.com/sustainability).
Fiber Optic Deployment: The need for high-bandwidth, low-latency data transfer between AI models, data storage, and end-users is driving continued investment in fiber optic networks globally. Submarine cables and terrestrial fiber backbones are critical components (source: telegeography.com).
Semiconductor Demand: The foundational hardware for AI, particularly Graphics Processing Units (GPUs) and specialized AI accelerators, is experiencing unprecedented demand. This drives significant capital expenditure by semiconductor manufacturers and their equipment suppliers, as seen in ASML's strong bookings and SK Hynix's capex boost (source: marketwatch.com, news item 6).
Investment Flows: Private market investors are increasingly targeting digital infrastructure. For instance, in 2023, global digital infrastructure investment reached record levels, with a significant portion directed towards data centers and fiber networks (source: preqin.com, author's assumption for specific figure, but trend is verifiable). This reflects confidence in the long-term growth trajectory driven by AI.
Scenarios
Three plausible scenarios outline the future adaptation of infrastructure to the AI revolution:
Scenario 1: Rapid, Coordinated Adaptation (Probability: 40%)
In this optimistic scenario, governments, private industry, and international bodies recognize the urgency and collaborate effectively to address the infrastructure demands of AI. Policy frameworks are swiftly updated to incentivize sustainable data center development, streamline permitting processes, and accelerate renewable energy deployment. Public-private partnerships flourish, channeling significant capital into grid modernization, new energy generation, and advanced digital networks. Research and development in energy-efficient AI hardware and cooling technologies receive substantial funding, leading to breakthroughs that mitigate environmental impacts. Regulatory bodies adopt agile approaches, fostering innovation while ensuring responsible development. Supply chains for critical components (e.g., chips, cooling systems) are diversified and strengthened through international cooperation. This scenario leads to a relatively smooth and sustainable integration of AI into the global economy, with infrastructure evolving at a pace commensurate with AI's advancements. Economic growth is strong, driven by AI innovation, and environmental concerns are actively managed.
Scenario 2: Fragmented, Reactive Adaptation (Probability: 50%)
This is the most probable scenario, characterized by uneven progress and significant bottlenecks. Infrastructure development lags behind the rapid advancements in AI due to a combination of factors: slow regulatory responses, fragmented policy approaches across jurisdictions, NIMBYism (Not In My Backyard) hindering data center and energy project approvals, and insufficient coordination between energy providers and digital infrastructure developers. Funding gaps emerge in critical areas, particularly for grid upgrades and long-duration energy storage. Environmental concerns, such as energy consumption and water usage, become more pronounced as infrastructure struggles to keep pace with sustainable solutions. Supply chain vulnerabilities persist, leading to delays and increased costs for AI hardware and data center components. This scenario results in localized infrastructure stress, higher operational costs for AI companies, and a slower, more disjointed realization of AI's full economic potential. Innovation may be stifled in regions with inadequate infrastructure, exacerbating digital divides.
Scenario 3: Stagnation/Significant Delay (Probability: 10%)
In this pessimistic scenario, major economic, geopolitical, or environmental shocks severely impede the necessary infrastructure adaptation. A global economic downturn reduces investment capital for large-scale projects. Geopolitical tensions disrupt critical supply chains for semiconductors and energy infrastructure components, leading to prolonged shortages. Regulatory overreach or a complete failure to establish coherent policies creates an environment of uncertainty, deterring private investment. Environmental concerns, particularly around energy and water, escalate into major public opposition, halting new data center and energy projects. Grid instability becomes widespread, leading to frequent outages that cripple AI operations. This scenario results in a significant slowdown or even stagnation of AI development and adoption. The economic benefits of AI are largely unrealized, and the world faces a prolonged period of technological and economic underperformance, potentially leading to social unrest and increased global inequality.
Timelines
Short-term (1-3 years): Immediate Data Center Build-out and Grid Stress. The immediate focus is on expanding existing data center capacity and initiating new builds, often in proximity to renewable energy sources or robust grid connections. Energy utilities face immediate pressure to manage increased demand, potentially leading to localized grid stability challenges and calls for urgent infrastructure upgrades. Policy discussions intensify regarding energy efficiency standards and permitting processes for digital infrastructure. Initial investments in advanced cooling technologies and localized renewable energy solutions for data centers become more common (source: industry reports, author's assumption).
Mid-term (3-7 years): Significant Investment Cycles and Evolving Regulatory Frameworks. This period will see substantial capital expenditure in new hyperscale data centers, significant upgrades to national and regional energy grids, and continued expansion of fiber optic networks. Governments will likely implement more comprehensive regulatory frameworks addressing AI's energy footprint, water usage, and land planning. Public-private partnerships will become crucial for financing large-scale, complex infrastructure projects. The development of specialized AI chips and more energy-efficient computing architectures will gain momentum, influencing data center design (source: worldbank.org, iea.org, author's assumption).
Long-term (7-15 years): Integrated Sustainable AI Infrastructure and New Urban Planning. By this stage, AI infrastructure is expected to be deeply integrated into national and global energy and digital ecosystems. A significant portion of AI compute will be powered by dedicated renewable energy sources, potentially with advanced battery storage or green hydrogen solutions. Smart grid technologies will be commonplace, dynamically balancing supply and demand. Urban and regional planning will incorporate AI infrastructure considerations, including optimal locations for data centers that minimize environmental impact and maximize resource efficiency. New cooling technologies, potentially leveraging waste heat for district heating, will be more widespread. The focus will shift towards circular economy principles for data center hardware and infrastructure (source: un.org, author's assumption).
Quantified Ranges
While precise future figures are subject to significant variables, existing trends and projections offer quantified ranges for key areas:
Data Center Energy Consumption Growth: Projections indicate that global data center electricity consumption could grow from approximately 260-340 TWh in 2022 to over 1000 TWh by 2030, with AI being a primary driver (source: iea.org, author's interpretation of various reports). This represents a potential 3-4 fold increase within a decade.
Investment in Digital Infrastructure: Annual global investment in digital infrastructure (including data centers, fiber, towers) is estimated to be in the range of USD 200-300 billion, with a significant portion now directed towards AI-enabling assets. This figure is expected to grow, potentially reaching USD 400-500 billion annually by the late 2020s to meet AI demands (source: preqin.com, author's assumption for specific figures, but trend is verifiable).
Water Usage by Data Centers: A single hyperscale data center can consume millions of liters of water per day for cooling, equivalent to the daily consumption of tens of thousands of households (source: google.com/sustainability, microsoft.com/sustainability, author's interpretation of public data). This range will vary significantly based on cooling technology and climate.
Renewable Energy Integration: To meet the sustainability goals of major tech companies, the proportion of renewable energy powering data centers is targeted to reach 100% by 2030 for many leading firms (source: re100.org, company sustainability reports). This requires substantial investment in new renewable generation capacity, potentially hundreds of gigawatts globally over the next decade (source: iea.org, author's assumption for specific figure, but trend is verifiable).
Risks & Mitigations
Risks:
1. Energy Supply and Grid Stability: The exponential growth in AI's energy demand could overwhelm existing grids, leading to blackouts or brownouts, especially in regions with insufficient generation capacity or aging infrastructure. This risk is exacerbated by the intermittent nature of renewable energy sources if not adequately balanced with storage or baseload power (source: iea.org).
Mitigation: Aggressive investment in grid modernization, smart grid technologies, and diverse energy portfolios including renewables, nuclear, and flexible gas peaker plants. Development of long-duration energy storage solutions (e.g., advanced batteries, green hydrogen). Demand-side management programs for data centers.
2. Environmental Impact: High energy consumption leads to increased carbon emissions if powered by fossil fuels. Significant water usage for cooling exacerbates water stress in arid regions. E-waste from rapidly evolving AI hardware also poses a challenge (source: nature.com, epa.gov).
Mitigation: Prioritizing renewable energy sources for data centers. Implementing advanced, water-efficient cooling technologies (e.g., liquid cooling, air-side economizers). Developing circular economy principles for hardware, including recycling and refurbishment programs. Locating data centers in regions with abundant renewable energy and water resources.
3. Regulatory Lag and Fragmentation: The rapid pace of AI development often outstrips the ability of regulators to establish comprehensive and harmonized policies, leading to uncertainty, inconsistent standards, and potential barriers to infrastructure development (source: ec.europa.eu, author's assumption).
Mitigation: Fostering international regulatory cooperation and best practice sharing. Adopting agile regulatory frameworks that can adapt to technological changes. Engaging industry stakeholders in policy development to ensure practical and effective solutions.
4. Funding Gaps and Investment Barriers: The sheer scale of investment required for AI infrastructure may outstrip available capital, particularly for public sector components like grid upgrades. High upfront costs, long payback periods, and perceived risks can deter private investors (source: worldbank.org, author's assumption).
Mitigation: Utilizing public-private partnerships (PPPs) to share risks and leverage private capital. Implementing targeted government incentives (e.g., tax breaks, subsidies) for sustainable AI infrastructure. Developing innovative financing mechanisms, including green bonds and blended finance.
5. Supply Chain Dependencies and Geopolitical Risks: Reliance on a few key regions or manufacturers for critical components (e.g., advanced semiconductors, specialized cooling equipment) creates vulnerabilities to geopolitical tensions, trade disputes, or natural disasters (source: marketwatch.com, author's assumption).
Mitigation: Diversifying supply chains and fostering domestic manufacturing capabilities where feasible. Strategic stockpiling of critical components. International agreements to ensure the free flow of essential goods and technologies.
6. Cybersecurity Risks: The concentration of vast amounts of data and computational power in AI infrastructure creates attractive targets for cyberattacks, potentially leading to data breaches, service disruptions, or intellectual property theft (source: enisa.europa.eu, author's assumption).
Mitigation: Implementing robust, multi-layered cybersecurity protocols, including AI-powered threat detection. Regular security audits and penetration testing. Fostering a culture of cybersecurity awareness and continuous training for personnel.
Sector/Region Impacts
Sector Impacts:
Energy Sector: Will experience unprecedented demand growth, necessitating massive investments in generation (especially renewables), transmission, and distribution infrastructure. Utilities will need to become more agile and integrate smart grid technologies. The energy transition will be significantly influenced by AI's demands (source: iea.org).
Real Estate and Construction: A boom in data center construction will drive demand for specialized industrial land, skilled labor, and advanced building materials. This will impact urban planning, land use policies, and potentially lead to localized real estate price increases (source: infrastructureinvestor.com, author's assumption).
Telecommunications: Continued and accelerated investment in high-bandwidth, low-latency fiber optic networks (terrestrial and submarine) and advanced wireless technologies (5G, 6G) will be critical. The telecom sector will be a foundational enabler of distributed AI (source: gsma.com, itu.int).
Manufacturing: Increased demand for specialized hardware (AI chips, servers), cooling systems, power infrastructure components (transformers, switchgear), and renewable energy equipment will boost the manufacturing sector globally (source: marketwatch.com, author's assumption).
Financial Services: Infrastructure funds, private equity, and institutional investors will see significant opportunities in financing digital and energy infrastructure. New financial products and risk assessment models will emerge to support these investments (source: preqin.com, brookfield.com).
Region Impacts:
Established Tech Hubs (e.g., Silicon Valley, Northern Virginia, Dublin, Singapore): These regions will continue to be centers of AI infrastructure development but will face increasing challenges related to energy supply, land availability, and environmental impact. They will need to innovate in efficiency and sustainability (source: datacenterdynamics.com, author's assumption).
Regions with Abundant Renewable Energy (e.g., Nordics, Pacific Northwest, parts of Latin America/Africa): These areas are becoming increasingly attractive for data center development due to access to cheap, clean power, potentially leading to new economic development and job creation (source: iea.org, author's assumption).
Emerging Markets: AI infrastructure presents both challenges and opportunities. While some may struggle with funding and regulatory capacity, others could leapfrog older technologies by directly investing in modern, sustainable AI-ready infrastructure, potentially attracting foreign direct investment (source: worldbank.org, author's assumption).
Geopolitically Strategic Regions: Control over critical infrastructure, particularly semiconductor manufacturing and key data routes, will become even more strategically important, influencing international relations and trade policies (source: author's assumption).
Recommendations & Outlook
For governments, agencies, CFOs, and boards, navigating the AI infrastructure revolution requires a proactive and strategic approach. The outlook for those who adapt effectively is one of sustained economic growth and technological leadership, while those who lag risk falling behind.
Recommendations for Governments and Agencies:
1. Develop Integrated National AI Infrastructure Strategies: Create comprehensive plans that align digital infrastructure, energy policy, land use planning, and environmental regulations. This should include clear permitting processes and incentives for sustainable development (scenario-based assumption: such strategies will accelerate adaptation).
2. Invest in Grid Modernization and Renewable Energy: Prioritize and fund upgrades to national energy grids to enhance capacity, resilience, and smart capabilities. Actively promote and incentivize the development of dedicated renewable energy sources for AI infrastructure (scenario-based assumption: robust energy infrastructure is foundational to AI growth).
3. Foster Public-Private Partnerships (PPPs): Establish clear frameworks and mechanisms for PPPs to de-risk and accelerate investment in critical AI-enabling infrastructure, leveraging private capital and expertise (scenario-based assumption: PPPs are essential to bridge funding gaps).
4. Promote Research & Development in Efficiency: Fund and incentivize R&D into energy-efficient AI algorithms, hardware, and advanced cooling technologies to mitigate environmental impact (scenario-based assumption: technological innovation is key to sustainable growth).
5. Address Skills Gaps: Invest in education and training programs to develop the workforce required for building, operating, and maintaining advanced AI infrastructure (scenario-based assumption: a skilled workforce is critical for successful implementation).
Recommendations for Large-Cap Industry Actors (CFOs, Boards):
1. Integrate AI Infrastructure into Long-term Strategic Planning: Recognize AI infrastructure as a core strategic asset, not just an operational cost. Factor its demands into capital expenditure planning, site selection, and supply chain management (scenario-based assumption: proactive planning will yield competitive advantage).
2. Prioritize Sustainability and ESG Factors: Invest in renewable energy procurement, water-efficient cooling, and circular economy practices for hardware. This mitigates regulatory and reputational risks and aligns with investor expectations (scenario-based assumption: sustainability will become a non-negotiable aspect of AI infrastructure).
3. Diversify Infrastructure Investments: Explore opportunities across the entire AI infrastructure stack, from data centers and fiber networks to renewable energy generation and storage solutions. Consider geographical diversification to mitigate regional risks (scenario-based assumption: diversification enhances resilience and captures growth opportunities).
4. Engage with Policy Makers: Actively participate in policy discussions to help shape effective and supportive regulatory environments for AI infrastructure development (scenario-based assumption: industry input is vital for practical policy outcomes).
5. Innovate in Design and Operations: Continuously seek out and implement cutting-edge technologies for energy efficiency, cooling, and modular data center design to optimize performance and reduce operational costs (scenario-based assumption: continuous innovation is necessary to remain competitive).
Outlook:
The AI revolution is poised to redefine the landscape of global infrastructure. The coming decade will be characterized by unprecedented investment in digital and energy infrastructure, driven by the insatiable demands of AI. We anticipate a significant shift towards more sustainable, resilient, and interconnected infrastructure ecosystems (scenario-based assumption). Regions and nations that proactively adapt their policies and invest strategically will likely emerge as leaders in the AI-driven economy, attracting talent and capital (scenario-based assumption). Conversely, those that fail to address these foundational infrastructure challenges risk falling behind, facing energy shortages, economic stagnation, and a widening digital divide (scenario-based assumption). The success of the AI revolution is inextricably linked to the success of its underlying infrastructure adaptation (scenario-based assumption).