What CERAWeek Made Clear: The Energy Transition Is Now an Execution Story

Against the backdrop of escalating geopolitical tensions and energy volatility, I left CERAWeek optimistic about the future of the energy landscape. The reason being that I believe the next phase of the energy transition will be shaped by a flywheel effect. AI-native software is starting to speed up planning and improve operations. Physical AI will increasingly help ease real bottlenecks in energy capacity development. And grid modernization will allow for a diverse energy system to function. These three themes build on each other and are self-reinforcing.

Conversations about energy diversification, resilience, and grid investment did not feel abstract compared to prior CERAWeek as many real-world commercial-scale results were presented. With a volatile global energy system and a structural demand for energy, in part driven by data centers, the question is no longer what needs to be built. It is whether the system can move faster and handle more complexity than it does today. 

Digital AI Is Starting to Matter as Traditional Workflows Struggle to Keep Up

I found that most conversations around AI at CERAWeek were rooted in the practical. The focus was on the application of AI within planning cycles, maintenance, downtime, asset performance, and forecasting. 

Even with electricity demand from data centers alone expected to rise sharply this decade, project timelines remain long and many core workflows still move too slowly. One AI platform serving generation owners said roughly half of operational performance loss in its sector is avoidable and that predictive maintenance tools could reduce false positives by up to 80%. Another platform is helping reduce utility planning and project timelines from six months to six weeks. AI that helps teams plan faster, catch problems earlier, and reduce avoidable downtime will have a material impact on project economics and system performance. The companies gaining traction seem to share a common profile: strong domain expertise, proprietary field data, and software that integrates with existing workflows. 

Physical AI Is Still Early, but Labor Problem Is Already Here

If digital AI helps with planning, physical AI will eventually help with the harder part: getting things built and maintained in the real world. The energy transition is a physical buildout story, not a digital one. More generation is needed, but so are transformers, transmission equipment, battery systems, and the people required to install, inspect, operate, and maintain them. Goldman Sachs estimates the power industry will need more than 750,000 new workers by 2030, which is why physical AI that alleviates labor constraints represents a massive opportunity.

Physical AI is getting more attention, even though it is still in its nascent stage, because the labor issue is real now. Field conditions are messy, dangerous, and hard to automate, and I believe progress will almost certainly be slower than the forecasts suggest. Early deployments in offshore wind inspection and solar material handling are showing real results and value even in narrow use cases. If physical AI can improve safety, reduce labor intensity, support maintenance, or turn field activity into better data, that matters in a sector where execution capacity is already a real constraint and inflationary. 

Grid Modernization Is What Makes Diversification Real

The third theme, which ties the flywheel together, is grid modernization. A major message at CERAWeek was that countries want a more diversified energy mix. Nuclear, geothermal, renewables, gas, and storage are all in the conversation for a balanced portfolio. That shift is partly about cost and reliability, but increasingly it is also about resilience. A concentrated energy system is now viewed as a strategic vulnerability. 

The challenge is that diversification is much easier to talk about than to operate. A broader generation mix creates more complexity, and legacy grid infrastructure was not built for that level of coordination. Estimates show the U.S. will need $720 billion in grid investment through 2030. Utilities still often rely on disconnected planning and operating systems that do not work together in real time. Without grid modernization, adding more generation sources can create coordination problems faster than adding usable capacity. 

Bottom line

My main takeaway from CERAWeek is that the next phase of the energy transition will not be won by any single technology. It will be won by how well the sector connects better planning, more efficient execution in the field, and a grid capable of handling a more complex generation mix. 

Photo credit: Pexels