Dr. Daniel Kearney, CTO at Firmus Technologies, driving sustainable energy-efficient computing innovation and tech leadership solutions.
Digital generated image of rectangular s haped data tunnel.
getty
As AI transforms from experimental technology to essential infrastructure, enterprises and nations face an unprecedented challenge: how to harness AI’s transformative power without breaking the electrical grid or compromising sustainability goals.
Let’s call this challenge “The Energy of Intelligence,” which is the concern about the growing energy consumption as AI advances. Solving this problem demands immediate attention and innovative solutions.
The AI Energy Imperative
AI has emerged as the defining technology of our era, promising exponential productivity gains and unlocking new economic value from limited resources. Countries now consider AI capabilities central to trade negotiations and competitive advantage, making AI not just a business tool but a sovereign requirement.
However, this AI revolution comes with a substantial energy cost. Global data center electricity consumption, driven largely by AI workloads, could more than double by 2030, reaching between over 945 terawatt-hours, rivaling the entire electricity consumption of Japan.
The challenge is particularly acute for AI training infrastructure, where rack are expected to require upwards of 600 kW of power by 2027, which, as Goldman Sachs puts it, is “50 times more power per rack than CPU datacenters of just five years ago” and equivalent to fitting “enough power for 500 US homes into the space of a filing cabinet.”
The mathematics are sobering. In the U.S alone, the data center sector could consume 9% of all grid power by 2030, up from 4% in 2023. While AI hardware energy efficiency improves by approximately 40% annually, these gains are consistently outstripped by exponential increases in deployed workloads and user adoption.
Three Laws For Sustainable AI Infrastructure
To address this challenge, we must embrace three fundamental principles—three “laws” that should govern how we build, deploy and scale AI systems globally.
1. The Law Of The Land: Sovereign AI Requirements
The first law recognizes that AI cannot be divorced from geography, governance and sovereignty.
Take Singapore, a 735.6 square kilometer island nation with limited natural resources but ambitious digital aspirations. With a median population age of 42.8 years and few natural resources, including power, land and water, Singapore faces the classic challenge of small nations in the AI era: how to build sufficient sovereign AI compute capacity within physical and energy constraints.
Similarly, Ireland illustrates how grid limitations can constrain AI ambitions. Data centers now consume 21% of Ireland’s electrical grid capacity, up from just 5% in 2015, leaving utility companies power-constrained and limiting their ability to offer additional compute capacity for AI workloads. This creates a direct threat to economic growth and competitiveness.
The Law of the Land demands that AI infrastructure be designed as fit for purpose to meet local requirements, respecting sovereignty requirements while operating within each nation’s unique resource constraints. This means moving beyond one-size-fits-all cloud solutions toward fit-for-purpose sustainable AI systems that can deliver value locally without overwhelming national infrastructure.
2. The Law Of Economics: Cost-Effective AI
The second law addresses the fundamental economic equation of AI: Systems must create more value than they consume. This principle extends beyond simple financial calculations to encompass energy efficiency, resource utilization and long-term sustainability.
Current AI scaling approaches often prioritize performance over efficiency, leading to systems that deliver impressive capabilities but at unsustainable costs. The Law of Economics demands a paradigm shift toward AI architectures that optimize for total cost of ownership, including energy consumption, infrastructure requirements and operational efficiency.
This economic lens forces critical questions: Does every AI application require the most powerful available model? Can we achieve 80% of the value with 20% of the energy consumption? How do we balance capability with sustainability to ensure AI remains economically viable at scale?
3. The Law Of Physics: Bandwidth-First, Latency-Tolerant Optimization
The third law reveals a strategic shift: AI systems need massive data highways, not lightning-fast responses.
Unlike traditional applications requiring instant responses, AI operations can tolerate delays—training waits minutes while applications like ChatGPT work perfectly, delivering responses over hundreds of milliseconds.
This creates an executive opportunity: Prioritize bandwidth over ultra-low latency for cost savings. Modern AI demands terabytes of data movement rather than microsecond response times, requiring sustained high-volume transfer over extended periods.
By designing infrastructure that maximizes data throughput instead of minimizing response time, organizations build more cost-effective, sustainable AI systems. This bandwidth-first approach enables distributed architectures positioned closer to business operations while maintaining competitive AI capabilities without premium speed optimization costs—critical as infrastructure spending scales enterprise-wide.
Breaking Infrastructure Breaking Points
The convergence of these three laws points toward a fundamental shift in how we approach AI infrastructure. Rather than accepting the current trajectory of exponentially increasing energy demands, we must embrace efficiency-first design principles that make AI sustainable at a global scale.
This means developing AI models specifically optimized for energy efficiency, building data centers designed for local deployment and sovereignty requirements and creating hybrid architectures that balance centralized training with distributed inference.
Companies must move beyond the “scale at any cost” mentality toward “scale with purpose” approaches that consider total environmental and economic impact.
The Path Forward
The Energy of Intelligence challenge is not insurmountable, but it requires immediate action and innovative thinking. Organizations must adopt the three laws as design principles, governments must create AI policy frameworks that incentivize efficient AI development and the technology industry must prioritize sustainability alongside capability.
The companies that will thrive in the AI era are those that recognize that true intelligence includes the wisdom to scale responsibly. They will build AI systems that enhance human capability while respecting planetary boundaries—systems that embody not just artificial intelligence, but intelligent sustainability.
As AI becomes increasingly central to economic competitiveness and national sovereignty, the Energy of Intelligence will determine which nations and enterprises can successfully harness this transformative technology with an enduring and sustainable future.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?