The electricity grid is arguably the best example of building the plane while you fly it. The grid’s transition from centralized power production to integrated more distributed rThe electricity grid is arguably the best example of building the plane while you fly it. The grid’s transition from centralized power production to integrated more distributed renewables, and most of those new renewable sources of power are intermittent and variable. This necessitates upgrades to the transmission and distribution systems and a rethink of how we plan and operate grids. Oh, and all these changes need to happen while the grid continues to deliver electricity to customers. Making it worse is the fact that the utility industry is — and please take no offense to this if you work in it — one of the slowest innovators of all industries I’ve worked with during my time at Lux.
Slowing this process even more is that throughout this transition, the grid needs to add significant capacity. The transportation sectors is decarbonizing via electric vehicles; the built environment is transitioning to heat pumps, and the production of fuels and chemicals might come via electrochemical routes that electrolyze hydrogen and CO2. Interestingly, of all these topics, short-term concerns over data centers emerged front and center in 2024 for Lux’s utility clients. In part, due to the emergence of consumer-facing generative AI solutions like OpenAI’s ChatGPT, data centers are being built at a rapid rate, faster than many utilities can keep up with.
If every problem is just an opportunity in disguise, then data centers are certainly an innovation opportunity but not only just due to the size of the challenge. Data centers are relatively flexible in their location, meaning they can be sited near nuclear power plants or geothermal hotspots. They also face significantly lower regulatory barriers. Compared to grid-connected projects that inevitably end up in a years-long interconnection queue and may need regulatory approval, data centers can employ assets behind the meter and avoid public utility commissions or regulations that restrict certain technologies from being tied to the grid. The challenge of meeting new loads, these unique characteristics, and the relative wealth of the tech sector make data centers an innovation opportunity.
Lux sees such opportunities in a few areas:
- Load flexibility and efficiency: While load management isn’t the most exciting of opportunities in climate tech, data centers can provide some flexibility to the grid. Not all processes are critical, and some can be timed to run at certain times when grid congestion is high. For example, processes used to train AI models are some of the most flexible loads. Additionally, significant efforts have focused on improving cooling efficiency, which is a significant load within data centers. Tapping into flexibility and focusing on efficiency improvements are extremely near-term focus areas able to make an impact on data centers without introducing new technologies in the center itself.
- Long-duration storage: When I was leading our coverage of energy storage about 8 years ago at Lux, nearly every long-duration storage company we spoke with was targeting grid applications. This is somewhat intuitive, as these applications will be needed in grids with high rates of renewables. However, few grids today have a level of renewables that need more than four hours of storage. Furthermore, projects involving the grid can face significant delays due to regulatory barriers. In data centers, storage would be behind the meter and could avoid interconnection queues. From the perspective of a startup developing these technologies, data centers and the deep-pocketed companies building them present a promising initial target market today.
- Waste heat utilization: Low-carbon heat has been one of the hottest topics at Lux in 2024, as many industries that use steam or other forms of heat are searching for low-carbon alternatives to combustion. Conveniently, one of the challenges of lowering energy consumption in data centers is how much heat they produce, given that about 40% of energy consumption goes toward air conditioning systems. Data center owners have options to use this waste heat, which could be converted back into electricity via organic Rankine cycles or Stirling engines or injected into district heating systems where available.
While utilities are quickly figuring out how to incorporate load growth from data centers, tech companies have the most to lose. Success in the broader AI space for these companies is required to sustain their lofty valuations, and that can’t happen without also growing the number of data centers. The solutions noted above are key to the energy transition even before considering data center impact. Utilities should look to leverage their urgency to trial novel solutions needed in the energy transition, viewing data centers as innovation hubs for the rest of the decade.