AI to Solve its Own Energy Challenge

The rapid growth of artificial intelligence is reshaping global energy landscapes, creating both unprecedented challenges and opportunities. As nations compete for AI supremacy, a truth emerges; the ability to power AI infrastructure at scale will determine competitiveness, but success hinges equally on leveraging AI itself to optimize energy systems. Here's how these dual imperatives intersect and what it means for the future of tech and energy collaboration.

The Need for Creative Solutions

Global energy demand from data centers is projected to double by 2030 (to 945 TWh, surpassing Japan’s current annual consumption) with AI workloads driving much of this growth (1,4). In the U.S., data centers could consume more energy by 2030 than the entire manufacturing sector (steel and iron, cement and concrete, chemicals and minerals processing.(5) The acute need for creative solutions stems from:

• Fluctuating energy profiles within data centers: AI training workloads require sustained, high-power usage over long periods, while inference (to generate output) cause sharp, irregular spikes in power consumption.(3) Inference has surpassed training AI models, further driving energy demand.

• Infrastructure bottlenecks: Utility interconnections often take years to secure, delaying projects.(3,7) Building bridges with energy solutions that shorten time to market for data centers, and creating new partnerships between utility providers and on-site energy providers.

• Geographic concentration: 50% of U.S. data center growth clusters in five regions, straining local grids.(7) There is a need to introduce solutions to open up markets and unlocking developments in regions not previously considered.

AI as Both Problem and Solution

While the idea that “AI at scale depends on energy at scale” may be true, there is a more nuanced perspective; AI must also become the architect of its own energy sustainability.

The Two Sides of AI

Innovating for Efficiency

While energy availability is critical, AI leadership also depends on its own innovation, and cross-sector collaboration is extremely important at this point. We should pay attention to overdevelopment and to prematurely obsolete infrastructure caused by today’s spiraling urgency. This is by far a not a new phenomenon. We have seen the same pattern in compute and data storage since the 1950’s with direct and supporting infrastructure demanding excessive expansion before the technologies became smarter and more efficient. We should prepare for grid interconnection of behind-the-meter energy systems where possible, in case of on-site energy generation in excess, and rapid innovation and smarter technology. Examples of energy efficiencies in the works:

• Chip efficiency: Next-gen processors like NVIDIA’s H200 and Google’s TPU v5 cut energy per computation by 30-50%, easing power demands.(6) 

• Workload optimization: Tools like Meta’s Llama 3 reduce model training energy by 75% through algorithmic refinements.(6)

• Industrial symbiosis: Brave Industries’ energy platform allows co-location of data centers and micro-grids, creating efficiencies in energy use.(3,8).

Building the Bridge Between Tech and Energy

There are plenty of experts in both camps, and as the energy industry is doing “its best” to meet the growing and evolving demands of the data centers, this moment calls for deeper understanding and partnerships between the tech and energy industries. The recent report by IEA (1) underscores the delay of data center projects due to grid constraints, reinforcing the urgency for collaboration. Key steps include:

1. Unified standards: Develop AI-specific energy metrics (e.g., kW/FLOP) to align tech and utility planning. (1,6)

2. Policy innovation: Fast-track permits for AI-optimized micro-grids and onsite nuclear/SMR deployments.(3,7,8).

3. Knowledge sharing: Forums like Infrastructure Masons are expanding to include utility and independent energy providers, infrastructure developers and chipmakers in the same rooms. Calling for more workshops for better public-private partnerships. (9)

4. Interdisciplinary leadership: Co-designing infrastructure, and prioritizing R&D for models that advance AI capabilities to optimize energy systems.

The regions and nations that will do well as AI continues to demand massive amounts of energy, will be those that not only generate more energy from alternative sources and strengthen their grids, but those who use AI itself to reinvent their energy systems. By uniting tech innovation with energy expertise, we can transform AI’s power crisis into a catalyst for sustainable progress, proving that the industry’s greatest challenge may also be its most powerful solution.

Sources

Deloitte. "As generative AI asks for more power, data centers seek more sustainable solutions." Deloitte Insights, December 12, 2024.

RPower. "Navigating the Power Demands of AI-Driven Data Centers – Challenges and Opportunities." RPower, June 1, 2024.

Google Blog. "A new approach to data center and clean energy growth." Google, December 10, 2024.

(1)Data Center Frontier. "New IEA Report Contrasts Energy Bottlenecks with Opportunities for AI and Data Center Growth." Data Center Frontier, April 23, 2025. https://www.datacenterfrontier.com/machine-learning/article/55285268/new-iea-report-contrasts-energy-bottlenecks-with-opportunities-for-ai-and-data-center-growth

(2) https://techxplore.com/news/2025-04-ai-surge-center-electricity-demand.html

(3) https://www.datacenterfrontier.com/energy/article/55268513/perspective-can-we-solve-the-ai-data-center-power-crisis-with-microgrids

(4) https://www.datacenterdynamics.com/en/news/iea-data-center-energy-consumption-set-to-double-by-2030-to-945twh/

(5) Reccessary. "Three key takeaways from IEA's first Energy and AI report." Reccessary, April 11, 2025.  https://www.reccessary.com/en/news/world-market/iea-energy-and-ai-report

(6) https://www.nature.com/articles/d41586-025-01113-z

https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works

(7) https://ourtake.bakerbotts.com/post/102k8yl/anticipating-the-ai-energy-impact-iea-highlights-growing-demand-from-data-center

(8) https://www.brave.industries/