Energy Industry Ramps Up Efforts to Solve Data Center Power Shortage

At the Baker Hughes Annual Meeting in Florence, Italy, energy leaders gathered to address a growing concern: the rising power demands of AI-driven data centers and their impact on infrastructure. With a Goldman Sachs projection indicating a 165% increase in data center power demand by 2030, industry experts debated solutions to avoid grid overload while ensuring reliable energy supply.

AI’s Soaring Energy Demand and Grid Challenges
The rapid expansion of generative AI and cloud computing has placed unprecedented strain on local power grids. Jeffrey Tapley, COO of Digital Realty, highlighted how AI has supercharged data center demand, with some markets seeing data center space prices double. However, power constraints and delays in grid connections—sometimes extending beyond a year—are causing major hurdles for new facilities.

“I don’t find comfort in being told that I’ll have power for a new data center by 2030,” Tapley said. “We prefer to look at how we can get power to the site ourselves.”

Shifting Site Selection and Onsite Power Generation
With hyperscalers demanding massive power near AI development hubs, site selection is becoming increasingly complex. While metro areas remain a safe bet for long-term cloud expansion, AI-focused data centers may have shorter lifespans, making long-term power investments riskier.

Tapley noted that renewables and battery storage aren’t always viable due to land constraints, while natural gas remains a preferred option—if infrastructure is already in place. However, extending gas pipelines to remote locations remains costly and time-consuming.

The Case for Onsite Energy Solutions
Jonathan Maxwell, CEO of Sustainable Development Capital LLP (SDCL), argued that onsite power generation is key to overcoming energy inefficiencies. He emphasized that a significant portion of energy is lost through transmission and conversion, making local energy generation more cost-effective.

“By the time data center operators pay for expensive power, most of it is wasted,” Maxwell said.

SDCL is increasing its investment in onsite energy for data centers, with batteries and natural gas generation emerging as practical solutions to replace traditional diesel backup generators. Additionally, waste heat from data centers could be repurposed for heating and cooling, improving energy efficiency.

Former U.S. Energy Secretary Dan Brouillette echoed concerns about grid capacity shortages, warning that the U.S. could face a 30 GW power shortfall within five years—and that was before the latest AI-driven surge in demand.

His takeaway? Data center operators will increasingly need to secure their own power sources rather than relying on local grids.

“Colocation will be more and more the future,” said Brouillette. “Bring your own power is going to be a major trend.”

However, while renewables are part of the long-term solution, land constraints and intermittency issues mean they won’t provide an immediate fix. Nuclear energy faces similar challenges due to long permitting timelines.

AI Efficiency Breakthroughs: Game-Changer or Temporary Fix?
A potential disruptor entered the conversation with China’s recent unveiling of DeepSeek AI, a breakthrough promising 90% lower energy consumption for AI workloads. While this innovation could slow the rate of energy demand growth, Brouillette believes it won’t change the long-term trajectory.

“Even if DeepSeek’s energy promises are proved out, it will impact the timeline of energy predictions but not the amount,” he said.

Looking Ahead
The AI revolution is pushing the energy sector to adapt at an unprecedented pace. Whether through onsite power generation, grid upgrades, or new efficiency breakthroughs, the industry must act now to prevent a power crisis that could stall digital transformation.

With data center operators increasingly exploring hybrid energy solutions, the next decade will be defined by innovation in power generation, storage, and efficiency technologies—reshaping how AI, cloud computing, and infrastructure intersect.