How AI and the Energy Transition are Redefining Real Estate Value
Understanding how the energy problems posed by AI could reshape the built world
Today’s Thesis Driven is a guest letter from Jenny Song, Partner at Navitas Capital, an early stage venture firm that has been focused on the impact of AI and the energy transition on the built world for over 15 years.
Every major technology revolution has redefined real estate value. From the Agricultural Revolution to the Information Age, technology developments have changed our relationship with land and buildings. Our view is that the Artificial Intelligence Era will do the same: real estate in the age of AI will be increasingly defined by energy and its availability.
Of course, AI will be applied to real estate operations directly. At Navitas, we spend a lot of our time examining technology’s direct use cases for the industry – for example investing in companies like EliseAI, a conversational AI applied to the housing sector.
But we are also closely watching the indirect, second-order effects. Airbnb’s initial impact, for example, was making room-sharing between private individuals easier. Since then, it reshaped multiple real estate asset classes from traditional hospitality to multifamily to single family home building.
In the AI Era, real estate may be increasingly defined by its relationship with energy production and transmission. Considering the energy intensity of AI data centers–as much as 50 times as intensive as a similarly-sized office building–and constraints on our ability to produce or transmit energy fast enough to keep pace, we see a world in which energy is increasingly constrained -- making its availability a more important and deterministic factor in all real estate development.
In today’s Thesis Driven, we’ll explain why this is the case and what it means, covering:
A framework for how to think about predicting real estate value at a macro level
The energy problem posed by AI and the rapid build of new data centers
How AI is converging with other mega-trends that are creating immense energy challenges
Why what we are doing today is not working to solve the problem fast enough
What this all has to do with real estate value and what it means for real estate developers, owners, investors and others
Location, location, location
No mantra is more obvious in real estate. But what makes a good location or a bad one? The answers are sometimes tautological – a good location is one where real estate values keep rising. How do these changes in real estate value start?
The differential value of land goes back to the Agricultural Revolution. Where land was fertile and water available for irrigation, communities of early humans were able to flourish and produce surplus resources and support surplus labor. These communities became villages, and then towns. Where towns could connect via waterways and trade their surpluses, they became richer and larger. This clustering of resources and access to trading those resources produced highly valuable real estate in port towns. It’s no accident that some of the most economically important and highest rent cities in the world today got their beginnings as ports.
The rise and fall of the Steel Belt in the United States follows a similar pattern. After the Industrial Revolution, iron ore and coal became the most valuable economic resources. With railroads moving coal from Appalachia to the North and Midwest, steel production flourished close to iron ore deposits along the Great Lakes. An influx of labor also contributed to the cauldron – immigrants from Eastern Europe streamed into the area alongside African Americans following railroads to the North in the Great Migration. Some of the fastest appreciating real estate of the time was in Pittsburgh, Philadelphia, or St. Louis.
A century later, the Information Age decreased the importance of the movement of physical goods but increased the importance of networks of talent and capital. The first semiconductor businesses located in Silicon Valley to be close to customers like the US Navy and NASA. Soon, engineering talent began to cluster in the area as did creative financial firms like venture capital and venture bankers. This not only led to a spike in real estate values in the region but also had major impacts on the type of real estate being built. With talent now the scarcest and most valuable resource, amenities rose in importance. Big Tech campuses outfitted with Michelin-starred restaurants and climbing gyms became the norm.
These history lessons illustrate how real estate has developed around resource constraints and mobility. The critical resources in each age varied – but in each case, real estate development and value clustered around the rarest and highest value resources, and along the routes by which to access those resources.
In the AI Era, our view is that labor and transit will diminish in importance. Instead, electrons will become paramount, with electron production and the physical constraints of the electrical grid shaping the future of development.
AI’s Energy Problem
AI has an insatiable appetite for energy. AI data centers are denser than normal data centers and consume more energy because of the chips themselves – GPUs which process more data and therefore consume more energy – as well as their higher cooling needs.
A data center consumes up to 50 times more energy per square foot compared with a commercial office. Training a model like Chat-GPT3 consumed as much energy as someone streaming Netflix for 185 years; An AI query is ten times as energy-intensive as a Google search.
While data centers make up just 2-3% of total energy use today, this is expected to double in just the next 4 years; JLL notes that construction has grown sevenfold in the last two years.
Convergence with other mega-trends
It’s important to set this increased strain on energy needs in the context of multiple other mega-trends converging at once. On the supply side, we are in the midst of a massive energy transition, shifting toward a cleaner, more renewable electric grid. Most states have passed mandates that require grid operators to hit certain thresholds of renewable or clean power; 23 of those states have a goal of 100% with deadlines ranging from 2030 to 2050. As a result, states are actively retiring "dirty" plants fueled by coal and natural gas and replacing them with renewables.
But there are other electricity demand drivers rising at the same time as AI data centers consume more power:
American Manufacturing. A massive reshoring effort is ongoing in manufacturing, much of which is driven by the energy transition – for example, battery and solar panel production. Compared with Industrialization 1.0, these facilities are lower in labor and higher in energy intensity due to significant process automation.
Electrified Transport. Whether cars or freight, a larger percentage of vehicles each year are electric, fueling through new charging stations and infrastructure instead of with fossil fuels. Less than 1% of vehicles on the road are currently electric, but that is expected to grow to 10-11% by 2030.
Building Electrification. From homes to commercial buildings, heat pumps, heat pump water heaters, and electric stoves are gaining share relative to gas furnaces, boilers and gas stoves.
Together with AI, these four major power demand drivers are all shifting work from a diversity of forms of "energy" – for example human labor in the case of work replaced by AI – to just one type: electrons on an electric grid.
To draw an analogy, imagine instead that we’re talking about vehicles on roads. This would be like going from a world with a multitude of transit types – cars, trucks, buses, subways, bikes, planes, railroads, ships, and even just walking – to one where you only have cars. If we lived in that new world, two things would happen: (1) we’d have to make a lot more cars. And (2) we’d have to build a LOT more roads and make them all wider. Even with those, the traffic would be so bad we probably wouldn’t leave our homes unless absolutely necessary.
This is exactly what needs to happen to electrons and the grid. We are going to have to produce more power than ever before. And we are going to have to build a lot more transmission to move that power to where it is needed.
We are not building fast enough
Utilities are now faced with the challenge of simultaneously closing old, dirty power plants while meeting increased energy demand. While projections for increased energy demand are not drastic in real terms, they are getting revised higher every year. In the context of historically flat production, doubling or quadrupling energy production and transmission is a daunting proposition.
The desire to build is there: power developers have more proposed projects in queue than ever before; more than the existing capacity of the US. Most of this is renewable and much of it includes battery storage. While this is encouraging, fewer than 1 in 5 projects in the queue actually get built. Most projects sit in the queue waiting for approval for years and never start construction because they can’t get permitted or interconnected.
Transmission
While an increasing need for power generation demands more transmission lines than ever before, we’ve actually been building fewer lines each year. This is creating pain across the board: for the power projects that can’t get interconnected as well as for the data centers and other energy consumers that are also starting to wait years for interconnection.
Regulation
The Federal Energy Regulatory Commission produced a new rule last year that aims to reduce queue congestion by introducing a "clustering" study process and mandating that utilities increase transparency around energy availability. Many experts, however, expect the impact to be limited. For one, many states are already suing; in addition, several operators like the California Independent System Operator already use a similar process but have not seen queues lessen.
Nuclear
Another "obvious" solution – but potentially a red herring – is nuclear energy. Some vocal technology leaders like Bill Gates and Sam Altman have put massive investment behind the new generation of nuclear reactors, also known as SMRs (small modular reactors). And public sentiment appears to be shifting more favorably toward nuclear, which could help regulators ease restrictions. However, new SMRs could still be more than a decade away from real impact on the grid – despite very publicly breaking ground on construction, for example, TerraPower (a Gates-founded company) has been at it since 2008 and still hasn’t secured a license for an initial demonstration reactor (construction started on non-reactor elements). Even if TerraPower or others are able to run a demo reactor safely, the leap to scale will prove to be another challenge: a scale supply chain does not yet exist.
There is unlikely to be a magic bullet. We are going to need to build a massive amount of new infrastructure, and until it is built energy will continue to be a key constraint. The US transcontinental railroad took just 6 years to build. The US interstate highway system took nearly four decades. We’re not optimistic this new wave of energy infrastructure will be built on a faster timeline.
That means that in the AI Era, the availability of electrons is going to shape our built world.
Implications for real estate developers, investors, and owners
If electrons and the electric grid become the primary constraints and drivers of the AI era, what does it mean for real estate?
All new development, and many retrofits (such as solar installations, EV infrastructure, etc.), may wait longer for interconnection. Because all types of real estate compete with each other in the same queue, we could start to see new home development, for example, compete with new logistics centers for energy capacity and transmission. With delays and long timelines potentially flipping profitable projects to unprofitable ones, site selection will need to incorporate energy as a lens and risk factor.
The highest and best use of land will change depending on its energy attributes. A plot of land or even an existing building with more renewable potential and/or capacity to access power through its grid connectivity could have more potential for different development use cases. That means understanding those energy attributes is going to become critical in any kind of investment or valuation decision, whether you are a developer, investor, owner, lender, or have any other financial stake in the asset.
We believe microgrids are going to become more popular – and they won’t be so micro in the future. We are already seeing large developments co-located with power generators and batteries in an attempt to bypass the lengthy timelines required to get connected via a public utility. If microgrids gain traction, more and more wires will become "private" and will be operated and managed by private control centers. We could also see matchmaking between different types of assets with complementary energy profiles.
As venture investors, we of course see this as an opportunity for technology companies. Software companies like portfolio company Paces, for example, provide data and risk management around energy interconnectivity and can help owners or investors assess the energy attributes of various properties. Optiwatt and Gridium are other Navitas portfolio companies helping utilities manage increasingly distributed assets and helping real estate portfolio owners understand and optimize their energy use, respectively.
We’re eagerly watching how things play out. If we were real estate investors, we’d also probably be buying land right now. We leave that part, readers, to you.
—Jenny Song
Get in touch: Working on something interesting on this topic? Whether you’re a start-up, developer or real estate expert—feel free to reach out to Jenny.
Recommend Thesis Driven to your readers
A guest post by
|