Ex-CIO Of Dept of Energy and EPA Ann Dunkin: Why Leaders Must Weigh AI Energy Demands As A Core Business Risk

AI has upended a basic assumption of enterprise technology. Power can no longer be treated as cheap or assured, and as grid instability collides with always-on digital operations, energy has become a strategic risk landing squarely on the CIO’s desk.

Ann Dunkin, a Distinguished Professor at the Georgia Institute of Technology with deep experience across the public and private sectors, brings a rare vantage point to the question of energy risk. A four-time enterprise CIO, she previously served as CIO for both the U.S. Department of Energy and the Environmental Protection Agency, where she worked at the intersection of digital infrastructure, policy, and grid operations. That experience now informs her view that energy reliability has become a core technology leadership issue, not a background operational detail.

"Many CIOs have never thought about energy. That's changing fast. More organizations are exposed to energy risk, and leaders need to know what those risks are and who owns them," Dunkin said. She believes CIOs should start by rethinking their understanding of the grid as a cyber-physical system.

• Digital and physical: "The grid is not just a bunch of physical assets," she explained. "It’s physical and digital assets, and you need both the transmission lines and the data lines, along with the tools to optimize power and data flows, to make the system reliable." The concern reflects a core challenge in the power supply chain, as reliability increasingly hinges on software, data, and orchestration, not just steel and concrete.

• The queue conundrum: During her time at the DOE, Dunkin gained direct insight into the operational realities of the U.S. grid. She noted that while data center demand for power is growing, bringing new energy sources online is a gamble, forcing grid operators to adapt. "About 70% of projects fall out of the interconnection queue. So how do we ensure that we have enough power in the right place to meet demand in a timely fashion?"

Dunkin said CIOs need to move beyond tactical compliance and elevate energy into their strategic risk planning. In her advisory work, and as a board member of the Global Interconnection Group, she often sees organizations rely on checklists and narrow controls while missing broader exposure at the enterprise level. That gap creates a false sense of security, leaving leadership unprepared for disruptions that can directly affect operations, uptime, and business continuity.

• Compliance blind spot: "I don't know how many organizations I go into and they either don't have a risk register, or their risk register is at a very tactical level," said Dunkin. "Leaders need to focus on strategic risk and avoid letting compliance distract them from active threats."

• Beyond 24/7: As CIOs plan for 2026, many will need to expand their view of resilience. "Energy is becoming an operational and cybersecurity risk at the same time," Dunkin noted. "CIOs need to be asking not just how to keep systems up 24/7, but what a cyberattack on energy infrastructure would mean for their operations and what their utility providers are doing to mitigate that."

• Where and why: Her perspective on latency suggests an opportunity to rethink data center geography. Dunkin offered a nuanced view, distinguishing between niche applications where latency can be a factor—like analyzing X-rays at a hospital—and the majority of AI workloads, where she believes the constraints are often overstated. "When I use ChatGPT, I don't know where that data center is, and it doesn't matter," she said. "The time that ChatGPT uses for 'thinking' is far greater than the time the data requires to move to me. We just need to change some of our assumptions."

The insight opens the door to an alternative strategy for managing the heavy energy demands of enterprise-scale AI. "We can reduce the impact of data centers not only by building better LLMs and moving to smaller, more specialized models," she said, "but also by moving the data to where the power is, rather than the power to where the data is."

The takeaway is a call for shared ownership and clearer accountability, starting with a conversation many organizations have never had. "It’s often not clear who inside a company is actually thinking about energy risk," Dunkin concluded. "It might sit with a facilities manager by default, but that’s not enough. CIOs need to be talking across the organization, making sure someone owns the risk and that teams are working together to address it before it becomes a disruption."