AI is fundamentally energy-bound. With power grids operating near capacity, the expansion of AI is increasingly constrained by kilowatt availability rather than computational design.
Infrastructure developers, operators and investors understand that built infrastructure, particularly in the form of data centers, is a key requirement for realizing AI’s potential to transform the global economy. Contributing elements to commissioning more data centers include:
- Available real estate
- High-speed communications cabling
- Proximity to internet connectivity hubs
- Smooth and efficient logistics
- Proximity to water for cooling
However, the most significant enabler - currently considered a key constraint due to existing and forecasted shortages – is the availability of generated power.
The near future of AI is energy-constrained rather than compute-constrained
AI capabilities are already expanding beyond what existing infrastructure can consistently and affordably power. This will drive higher costs of compute, which may slow AI adoption and development.
However, the power supply availability constraint has high potential to drive investment towards efficient hardware and software with lower power demands. In turn, this is expected to continue reinforcing the current geopolitical advantage held by power-rich countries.
The Jevons Paradox, named after a 19th-century economist observing scale effects of coal availability on manufacturing, tells us that increased efficiency and lower costs of AI will drive higher rather than lower energy consumption. As the cost decreases, it is offset by an increase in usage. So far, evidence bears out this hypothesis in the current era, with AI soaking up ever-increasing amounts of energy.
Geopolitical disruptions to global energy supply flows
The current spikes in fuel prices resulting from geopolitical turmoil, particularly the ongoing conflict between Israel/US forces and Iran, present an additional cost-strain on AI.
Disruptive geopolitical events that affect energy supply flows compound the challenges of powering data centers with renewable energy. While renewable energy infrastructure and power generation exist, they are currently unable to match hourly electricity demands until battery storage technology is sufficiently developed. In this regard, the issue is twofold: building renewable energy production and distribution infrastructure, and developing improved battery storage technology, both require significant investment and time.
Energy constraints as a systemic risk for data center operators
Current insufficient global energy production capacity continues to convey geopolitical advantage to countries that can quickly build and scale new power plants. To preserve competitive advantage, hyperscalers and other data centers developers must be fully cognizant of regulatory and market forces affecting the availability of power at a given location. Furthermore, in Control Risks’ experience, data centers developers who treat power generation and distribution providers as key partners and stakeholders enjoy the greatest project success.
Both data centers and the power generation facilities they depend on are critical infrastructure and need to be physically and digitally protected. While competitive advantage rests in the developed world and wealthier countries, facilities in these countries are not immune to geopolitical disruptions or potential threat actors. Therefore, as new energy and data centers infrastructure facilities are developed, it is increasingly important for owners and asset design teams to factor in protective security considerations throughout the planning and design stages.
Security from the outset as a driver of resilience
Energy production facilities typically have high security requirements due to their designation as critical national infrastructure (CNI), and these are often mandated under government regulatory structures. However, physical protection requirements for data centers are less stringently regulated. In order to fulfill mandatory security requirements imposed by insurers, lenders or customers, data-centers developers, owners, investors, and operators are newly aware of the benefits of efficiency in ensuring effective levels of protection.
Our clients recognize that physical and digital vulnerability and risk reduction measures are most efficiently and effectively incorporated into new build designs for data centers, rather than as an afterthought or retrofit effort.
As the quantum and scale of data centers grows to meet increasing demand, so too will the financial and branding benefits derived from effectively and efficiently protecting AI infrastructure and enabling power facilities.
Article written by: Roberta Brezinski, Beth Cartier & Aaron Thatcher