There are numerous ways to invest in the Artificial Intelligence (AI) boom. Some high flying chip stocks like Nvidia have captured the spotlight, but there are more ways to invest in a broad wave of AI adoption than just semiconductor chips. Chips are essential, but without the infrastructure to support them they’re useless. Investors should also consider the sector powering the data centers needed for AI development.
According to the Electric Power Research Institute, data centers are forecasted to consume up to 9% of US electricity generation per year by 2030, up from 4% in 2023. That doesn’t sound like much, but between replacing old power generation and turning on new power arrays, the step up is significant. The reason is AI. The energy needs for AI computing are immense. To put the energy consumption into perspective, AI queries – just asking ChatGPT to clean up that email to your boss – can require approximately ten times the electricity of traditional internet searches. Training new models and running more complicated queries suck even more power.
The need for more power generation is so intense that it’s actually becoming the primary obstacle for companies looking to innovate in AI. The ideas are there, the computational chips are expensive, but they’re there, but you need to figure out how to meet the energy needs for the data center to function.
As this space continues to play out, we think it’s worth exploring opportunities within energy generation where investors can benefit from partnerships between both the upstream and downstream players in the AI market.
Data Centers Need Electricity…and lots of It
To explore the first-phase beneficiaries providing key inputs for AI advancements, it’s worth breaking down what data centers are used for when it comes to artificial intelligence. Data centers are the backbone of this AI boom as they store the hardware necessary to train AI models. As modern technology companies have grown larger and larger, and the amount of equipment needed to maintain operations and cloud services has skyrocketed, the energy demand from hyperscalers has followed suit.
Hyperscalers – companies that operate cloud computing infrastructure – provide the storage and compute power necessary to train AI and are faced with the difficult task of deciding where to source the energy needed for their daily operations. Major players within this space include Amazon Web Services, Microsoft Azure, and Google Cloud, all of which are forming partnerships with local utilities and power providers to ensure their energy demand is met.
With the sweeping transition to cloud data storage, electricity demand has turned into the biggest problem facing the industry. Hyperscalers providing infrastructure and platform services just can’t get enough of it, and the world of institutional finance is taking advantage of the opportunity. As we’ve seen recently, Blackrock teamed up with Microsoft (MSFT) and MGX (an AI fund out of The United Arab Emirates) to announce the launch of a $30 billion AI infrastructure investment fund focusing on building out data centers and energy infrastructure. While this may seem like a massive investment initially, it pales in comparison to the total investment in AI from the top 5 hyperscalers over the next few years. For context, current projections indicate the investment levels will reach a combined $1 trillion in 2027 according to S&P Global Market Intelligence.
Straight to the Source
The way power is generated and sold to large hyperscale consumers is going to change. Whereas in the past many data centers may have tapped into local electrical grids, the future of data center electrification involves dedicated power construction and long-term power purchase agreements (PPAs).
Hyperscalers prefer PPAs over market spot pricing since they’re able to guarantee stable energy prices for an extended period, typically over ten years. This level of certainty provides a hedge against unwanted price fluctuations that would affect hyperscalers’ bottom line and put upward pressure on already substantial model training costs. Perhaps more important though is the desirability of long-term contracts for developers and investors. PPAs from cash flush hyperscalers give developers the ability to secure financing for rapid construction of new power generation facilities. Knowing you have a reliable counter party allows for better profit forecasting! Furthermore, many of the power generation projects are dedicated solely to the data center and disconnected from the broader grid – a contract stipulating a minimum rate of return is essential for a project that takes years to break even and only has a single customer.
Finally, it’s worth pointing out that while renewables like wind and solar are the preferred long term source of power for the data centers, the demand right now is high enough that other options are on the table too. Nuclear power is back in the spotlight and natural gas turbines offer rapid deployment. Each power source comes with pros and cons – some are more flexible and some are harder to ramp up and down with data center demand. There will be a wide variety of solutions to the power conundrum over the next several years and many solutions that start off as off grid or dedicated power supplies may eventually end up becoming part of the larger power system.
Opportunities in Renewables
The long term demand for electricity is an opportunity to accelerate the build out of clean energy solutions. The top hyperscalers have set ambitious goals to reduce their carbon footprints – renewable power is a must for them over longer time frames. Among the top names, Google is targeting to operate its data centers on carbon-free energy by 2030, and Microsoft and Amazon both hope to shift to 100% renewable energy by 2025.
The massive undertaking is already underway as Microsoft (MSFT) and Brookfield Renewable Partners (BEP) signed the biggest-ever clean power deal this year for their data centers in the US and Europe, effectively penciling in 10.5 gigawatts of renewable energy capacity starting in 2026 and estimated to cost more than $10 billion. Keep in mind that 10.5 gigawatts is 3 times larger than the amount of electricity consumed by all data centers in Northern Virginia – commonly known as the data center capital of the world due to its favorable state tax incentives and best in-class access to power and internet connectivity.
With the steadfast increase in renewable energy demand due to corporate mandates, there are several ways to invest. We’ve already touched on the financing side where groups like Blackrock or Brookfield are raising capital to fund the construction of new power projects. Then there’s the manufacturing side, companies that make solar panels or electrical components, or companies who build wind turbines like GE Vernova (GEV). Finally, there are the utilities who want to own and operate the power generation facitlities over the next several decades, companies like NextEra Energy (NEE) or Constellation Energy Corp. (CEG).
Finally, there are cutting edge innovations in the power space that are less proven. As an example, innovative nuclear technology, like small modular reactors, is progressing rapidly and has received extensive backing from the US Department of Energy. These new reactors, built by firms such as NuScale Power (SMR), are smaller than traditional reactors and don’t have to be custom-built on-site. The process of generating electricity through nuclear fission is the same, but the benefits include reduced construction time, enhanced passive safety features, greater flexibility in deployment, and more efficient fuel usage. For broad adoption with data centers in the US, we’ll need to see continued investment in generation capacity and regulatory compliance, so patience is key.
Far from Being Over
The rapid expansion of artificial intelligence and pursuing energy demand are creating significant opportunities in the energy sector. As hyperscalers like Amazon, Google, and Microsoft ramp up their AI infrastructure, the need for massive amounts of electricity will drive investments in both traditional and renewable energy sources. While fossil fuels may provide immediate solutions due to existing infrastructure, the long-term focus is shifting toward renewables, such as wind, solar, and even nuclear power to meet sustainability goals.
Well-established and innovative companies in the energy sector, especially those forming strategic partnerships with hyperscalers can offer an alternative method to invest in the AI market besides buying chip companies or hyperscalers directly, even if the full societal benefits of investment end up taking years to play out. The ongoing collaboration between energy and AI firms will be crucial for supporting AI adoption and addressing challenges that lie ahead. Regardless of any short-term uncertainties, the unrelenting demand for energy remains far from being satisfied, and investors should continue to monitor the space for opportunities.
