The past month has seen a flurry of new power purchasing agreements (PPAs) between big tech and nuclear power companies—reviving the fortunes of a sector that had been struggling since the high-profile cancellation of Nuscale’s commercial pilot with Utah’s utilities and the cost overruns at Vogtle 3. Tech companies believe that they can gain major returns from investments in artificial intelligence technologies, and that means expanding their supply of compute capacity by building data centers. Data centers are power-hungry and run around the clock; intermittent energy sources like solar and wind won’t cut it. Nuclear remains one of the best, if not only, options for tech companies to secure large volumes of firm power while still meeting their decarbonization targets.

So, in the past month, Amazon, Google, and Microsoft have all signed power purchase agreements (PPAs) with nuclear developers by promising to purchase their output at a fixed price if they can come online. For tech firms, these PPAs let them secure power to their data centers ahead of time at a guaranteed price; securing power from these nuclear developers, even at a slightly more expensive price relative to other suppliers, is worthwhile to them because signing now is better than paying more later when load growth does increase and power is generally more expensive. They are also convinced the AI business is lucrative enough to justify the high input cost of locking in power prices in the range of $100/MWh or more More importantly, they do not have to put up money immediately: it’s an “if you build it they will come” situation. 

For a developer, having a PPA in hand helps it raise the capital it needs to build the asset since they now have a customer and cash flow that is guaranteed. However, if we zoom into the deals themselves, there are very significant differences between how the tech giants have structured their agreements. These illustrate the potentials and pitfalls of PPAs for catalyzing project finance investments across different types of technologies. 

Microsoft’s PPA is the most straightforward. Its contract with Constellation Energy purchases a portion of the output of a shuttered reactor at the Three Mile Island site in Pennsylvania. Here the PPA is particularly useful because it is financing the restart and operation of a particular, already well-established asset. The legacy reactor at Three Mile Island is a relatively bespoke piece of technology, but one with well-known economic and technical parameters. Assured demand is pretty much the most useful tool to get it running again.

The other deals are different because they are signed with Small Modular Reactor (SMR) developers. SMRs are a new, unproven technology that promise to be more efficient than current advanced reactors because they require less fuel and land than their conventional counterparts, like those at Three Mile Island. Most importantly, SMR manufacturers hope to achieve huge cost savings by turning nuclear reactors into manufactured products rather than bespoke projects. In the nuclear industry, the line between the designer and manufacturer of the reactor and the project developer is often blurred because of the scale and “bespoke-ness” of even the most standardized large reactor. But if SMRs really do become manufactured goods—where developers can replicate them quickly and keep parts in inventory—then it might make sense to separate the two functions as the industry scales. This is where things get complicated. because financing a new company that is doing production, but has yet to achieve standardization, sufficient piloting, nor scale is not the same as financing a project. A PPA can help  with commercial deployment, but it’s not a substitute for up-front capital to sink into building, testing, and ironing out a technology. This is where the first rounds of cost reduction will come from, and is also where the capital constraints are.

Google’s deal with Kairos power is interesting because it does more to target those constraints. Google is targeting multiple reactor sites built by the company. This is fairly unique because it is not tying a PPA to a single project and diversifying the company’s risks. However, as my colleague Advait has written, it still puts Google and Kairos in the difficult place of searching for financing for first-of-a-kind projects, even with the large amount of federally backed capital dedicated to deployment coming down the pipe. The Google deal is, thus, valuable to Kairos as it tries to create a pipeline of demonstration projects that can get their technology from first-of-a-kind to standardization and deployment at scale. 

This brings us to Amazon’s deal, which is more than a PPA. As part of its efforts, Amazon has invested $500 million into X-Energy, which is the company developing the actual reactor. It has also signed PPAs with two utilities, Energy Northwest, a public power joint purchasing entity in Washington State, and Dominion Energy in Virginia to install X-Energy’s reactors at new and existing nuclear sites. (This latter arrangement is worth paying special attention to because there are significant potential efficiencies from pairing SMRs with already-existing and new large scale generation on one secured site). Amazon is not only ensuring future cash flows to specific projects, but injecting low-cost capital into the production side of each project’s supply chain. Equity financing is critical to supporting a nascent manufactured-goods industry, providing X-Energy, as a supplier, an internal cash cushion with which to absorb the delays and difficulties that the first-of-a-kind projects Amazon’s PPAs supports will inevitably face. This also gives technologies time to achieve gains on the “learning curve” by spending the money to build out these technologies and put engineering concepts into practice. Without an equity cushion, project developers themselves have reservations about the capacity of new companies to deliver. Jason Ketchum, the CEO of utility holding company NextEra energy, recently explained that he doubted the economic viability of SMRs, in part, because “a lot of [SMR equipment manufacturers] are very strained financially and there are only a handful that have the capitalization that could carry them through the next few years.” 

Compared to its competitors, Amazon is modeling the right approach to developing SMRs that few private investors can replicate. It is rare to have a firm that, on one hand, has lots of free cash flow to invest in another risky company and, on the other, high volumes of energy demand. But we needn’t leave this strategy just to big firms who have expensive plans for what to do with that energy. The public sector, especially the Federal government,  can easily take inspiration from Amazon’s experiment in this coordinated nuclear financing strategy. Like Amazon, We should be providing upfront equity or grant financing to advanced energy technologies and we should be promising to buy their output and put it to use. This is a strategy we should apply to nuclear, long-duration storage, geothermal, carbon capture, desalinization, and a host of other technologies we feel could augment our energy needs. And we should ensure projects that achieve production can have an offtaker to reward their efforts.