One of the biggest questions around the boom in generative artificial intelligence has been how much energy will be needed to bring the full potential of the revolutionary—but incredibly power-intensive—technology to businesses and consumers.
As demand for generative AI continues to escalate and it becomes embedded into more products and services, energy demand is expected to intensify. Proprietary Morgan Stanley Research indicates that generative AI’s power demands will skyrocket 70% annually. By 2027, generative AI could use as much energy as Spain needed to power itself in 2022.
“We were surprised by our own projections showing such a rapid surge in power demand within just a few years,” says Simon Flannery, Morgan Stanley Research analyst covering North American telecom services and communications infrastructure. “We think the huge step up in energy needs is not well understood by the market and hasn’t been priced into a number of stocks.”
Generative AI’s massive appetite for energy could be a positive for data centers and power providers. And while the market has dour views on the technology’s sustainability profile, Morgan Stanley analysts think generative AI has the potential to deliver significant benefits toward decarbonization. Here’s what investors should consider.
Opportunities in Data Center Infrastructure
The reality of generative AI’s astonishing digital capabilities is that the enormous amounts of data required to create human-like speech, text, images and video often has to at some point run through physical hardware (at least for now). Data centers—giant warehouses where computing machines and hardware are stored—have seen tremendous uptick in demand and investment corresponding with generative AI’s rise. As a result, the industry is on track for significant growth. In fact, by 2025, generative AI’s power demands alone could equal a third of what data centers needed for all of 2022.
“For an industry with growth estimates that are typically around 10% per year, this incremental jump in power demand can clearly be an incremental driver of growth,” says Flannery.
In the short term, the growing need for more data centers to power generative AI apps and programs could create investment opportunities in the companies that build computing power infrastructure. These include integrated power-management systems providers and makers of factory-automation, air-conditioning and remote-monitoring systems. Additionally, property owners that specialize in industrial and logistics facilities (especially those with significant land banks or properties that can be easily converted to data center space); suppliers of cooling systems, racks, cables, processing units and IT management systems; as well as makers of flash storage systems, which handle 90% of current cloud storage needs, could also see boosts from generative AI.
Energizing Power Providers
Data center growth is heavily dependent on the expansion of the power grid, which currently is constrained by limited power line capacity; delays in planning and permitting for new transmission and distribution projects; and supply chain bottlenecks. In other words, power companies will need to develop significant new generation and transmission capacity to meet this surging demand.
Several segments of the power sector are likely to benefit from generative artificial intelligence growth.
Regulated utilities are the only companies that can provide power infrastructure at the massive scale needed to meet data center growth projections. Large-scale utilities could generate meaningful earnings from increased generative AI demand, although meeting the load requirements may require capital investments of $5 billion to $10 billion and could increase capital budgets for power generation by 10% to 20% annually.
Fuel-cell manufacturers that can provide base-load power to data centers will be in demand. Fuel cells are easy to deploy and can operate at a high capacity, offsetting the intermittency of most renewable power sources. If data centers are going to meet their power needs with renewable energy, fuel cells are key.
Nuclear power generators offer a unique opportunity to locate data centers on site and provide consistent, uninterrupted power without external connections. This would lower data center power costs by eliminating transmission and distribution charges—which can represent as much as half of a typical electric bill.
Providers of power purchase agreements (PPAs) work directly with large customers, usually from wind and solar generation, and enter into long-term agreements to provide electricity.
Interconnections, such as medium and high-voltage cables, are essential for balancing power demand and supply across regions, so the manufacturers of these products could see a boost as well.
“The location of data centers will also be key to individualutilities’ growth,” says Flannery. “For example, certain regions of the U.S. have dominated data center buildouts because of the availability of crucial resources like water for cooling, fiber optic networks for efficient connections and availability of land and real estate for expansion.”
Sustainability
While generative AI’s need for power could contribute to global carbon emissions the technology itself could drive long-term sustainability benefits through power grid optimization, smart agriculture, more precise weather forecasting and analyses of ocean impacts from climate change, and improving carbon capture and storage technologies.
And data center owners and operators have adopted decarbonization and net-zero targets and therefore hope to meet their electricity needs with renewable energy.
“We see a large portion of the incremental power needs for AI being sourced from zero or low-carbon technologies,” says Stephen Byrd, Morgan Stanley’s Global Head of Sustainability and Clean Tech Research. “The surge in AI-driven power demand could serve as an underappreciated driver for the manufacturers of large-scale and distributed clean energy technology and developers of wind, solar, energy storage and fuel cells.”
Quantifying what portion of generative AI’s future power consumption requires evaluating a wide range of variables, including different types of data centers with disparate power usage profiles and improved semiconductor performance over time. Even so, Morgan Stanley analysts believe that not only can generative AI power demands be met with sustainable sources, but its massive demand for power can also advance sustainable energy technology across sectors.
For deeper insights and analysis, ask your Morgan Stanley Representative or Financial Advisor for the full report, “Powering GenAI: How Much Power, and Who Benefits?” (Jan. 29, 2024)
