Beyond Chips: Investing in AI's Critical Foundational Infrastructure

Recent developments in AI have resulted in an unprecedented surge in computational demands. High-profile investments, such as Amazon's $20 billion commitment and various governmental initiatives like the US Executive Orders, have heightened the spotlight on energy, infrastructure, and connectivity systems. The foundational components of AI infrastructure, often overlooked in conventional investment models, now lie at the heart of AI scalability. As detailed in the research, the ecosystem comprises elements such as:

• Energy systems:
  With scenarios projecting US power demand of data centers scaling from 4 GW in 2024 to 123 GW in 2035.
• Cooling and thermal management systems:
  With new cooling technologies emerging as traditional methods become insufficient for high-performance workloads.
• Real Estate:
  Specialized compute real estate with optimal access to fiber optic networks and favorable climates.
• Networking:
  Resilient connectivity solutions ensuring high-bandwidth and low-latency communications.

These dynamics underscore why investments in overlooked infrastructure components offer substantial opportunities—and accompanying risks—that are critical to the sustainable growth of the AI sector.

The research set out to address the following key questions:

• What are the key non-semiconductor infrastructure components that present significant investment opportunities and risks in the AI build-out phase?
  This includes power grid upgrades, advanced cooling systems, specialized real estate investments, and expansive optical networking.
• How will escalating energy and water demands for AI data centers impact grid stability?
  The research investigates the influence on grid stability, the shift toward renewable energy sources, and the geographical reallocation of data center investments.
• What policy, regulatory, and geopolitical factors shape competition for AI infrastructure resources?
  In an era defined by data sovereignty, supply chain security, and national AI strategies, understanding these factors is pivotal to assessing long-term investment viability and regional leadership.

The report synthesizes a wide array of evolving industry data, projected technological shifts, and insights from policy reform initiatives.

A multi-source approach was adopted in this research, integrating quantitative and qualitative data from industry surveys, government reports, and academic insights. Key sources include:

• Industry Analyses by Deloitte and LandGate:
  Detailed projections on energy consumption, thermal load management, and permitting challenges.
• Governmental Action Plans:
  Insights from the Trump Administration’s AI Action Plan and federal proposals aimed at accelerating data center permitting and real estate development.
• Technical and Supply Chain Studies:
  Evaluations of advanced cooling methodologies, geothermal power potentials, and integration frameworks for supply chain resilience supported by real-time tracking and data analytics.
• Global Reports:
  Comparative studies by the World Economic Forum and international grid modernization analyses.

The research represents an integrative review of literature and data narratives that capture the multi-dimensional aspects of AI infrastructure investments.

Energy Demand and Grid Challenges

• Surging Power Demand:
  Deloitte’s analysis projects a dramatic increase in US AI data center power demand—from 4 GW in 2024 to 123 GW by 2035—with AI workloads anticipated to account for up to 70% of total data center energy demand (approaching 176 GW by 2035).
  Implication: Massive infrastructural upgrades are needed, and hyperscale projects may require individual capacities up to 2 GW. Facility expansions covering tens of thousands of acres demonstrate the scale of investment and echo urgent challenges in securing energy reliability.
• Grid Stability and Renewable Transition:
  US grid operators forecast new loads of 20 GW by 2030, contributing to 6.7%–12% of overall US power consumption by 2028. Initiatives like capacity auctions and demand response programs are being deployed to curtail interconnection delays.
  Geographical Impact: Regions with existing robust power grids and access to renewable energy have a competitive advantage, influencing location choices for future AI data centers.
• Permitting and Infrastructure Delays:
  Extended interconnection wait times—averaging up to 7 years—and delays in permitting (2+ years) represent significant bottlenecks. Over 11,000 power generation projects face permitting challenges, causing high withdrawal or cancellation rates.

Advanced Cooling, Thermal Management, and Power Solutions

• Thermal Loads and Cooling Technologies: Hyperscale data centers experience immense heat loads, where individual high-performance GPUs can generate up to 3 kW each, and rack-level heat production ranges between 5–50 kW.
  Conventional air cooling methods are increasingly inadequate in managing peak thermal outputs, driving the need for innovative cooling approaches.
• Emerging Cooling Solutions: Integration of direct-to-chip cooling, immersion cooling, and microconvective technologies is on the rise. These methods are being combined with advanced power management innovations such as Flex’s Vertical Power Modules to enhance energy efficiency and ensure system scalability.
• Energy-Efficient Cooling: Strategies such as liquid cooling, supplemented with methods like rainwater collection and geothermal cooling, are crucial.
  Enhanced Geothermal Systems (EGS): Can potentially meet up to 64% of hyperscale data center power demands by the early 2030s with cost optimizations of 31–45% over traditional deployments.

Real Estate Considerations and Site Selection

• Specialized AI Compute Real Estate:
  Real estate investments for AI infrastructure extend far beyond location selection. Key requirements include:
  • Proximity to robust fiber optic networks and internet exchange points
  • Cooler climates for natural thermal management
  • Secure, redundant power supplies with dedicated utility partnerships
• Innovative Site Selection Tools:
  Machine learning–driven site selection platforms are being developed using datasets such as the National Solar Radiation Database and satellite imagery.
  These tools are instrumental in identifying optimal sites that reduce environmental opposition while maximizing operational efficiencies.
• Regulatory and Zoning Incentives:
  Tailored zoning arrangements and relaxed permitting frameworks (as seen in the Trump Administration’s AI Action Plan) can accelerate the development of data centers, particularly on federal lands where environmental reviews are streamlined.

Networking and Data Connectivity

• Resilient High-Bandwidth Networks: An often-overlooked but critical aspect of AI infrastructure is the network’s ability to support massive volumes of data transfer. With optical networks and redundant high-speed interconnections, ensuring low-latency connectivity across distributed data centers is essential.
• Integrated Connectivity Strategies: Networking solutions are evolving to include on-site high-bandwidth connections and co-location with dedicated exchange points. This strategic integration minimizes data transfer bottlenecks and supports reliability in an era marked by latency-sensitive AI applications.

Policy, Regulatory, and Geopolitical Implications

• Policy Reforms and Federal Initiatives:
  The Trump AI Action Plan exemplifies efforts to expedite data center construction by streamlining permitting and scaling back certain environmental regulations. The aim is to balance rapid infrastructural buildout with strategic economic growth.
• Geopolitical and Data Sovereignty Considerations:
  Global data centers face additional regulatory hurdles pertaining to cross-border data flows and state-imposed cybersecurity laws. Frameworks like the EU’s GDPR and China’s Cybersecurity Law considerably influence the competitive landscape for AI infrastructure investments.
• Supply Chain Security and ESG Integration:
  Firms that successfully bolster their supply chain resilience through ESG measures not only improve production efficiencies (as seen in studies of Chinese A-share companies) but also alleviate financing constraints. Key KPIs such as “Days to Normalcy” and “Elasticity of Volumes” are indicative of how well companies adapt to both geopolitical and regulatory risks.