Every few generations, a new technology emerges that is so transformative and all-encompassing, it reshapes the entire economic system to its very foundation. And when this happens, the immense value created often spills over from the initial innovators themselves to a myriad of other industries and sectors that end up being revolutionized as a result.

Today, that technology is AI and to understand how it is reshaping industries and the opportunities being created in sectors like real estate, it’s helpful to first look back at previous periods in history where transformations of similar scale to what we are witnessing now have occurred.

Arguably the most apt comparison is to go all the way back to the late 18th century, when the world was in the initial stages of the First Industrial Revolution, thanks to James Watt and the advent of the steam engine. It wasn’t the steam engine alone, though, that changed everyday life but rather the second and third order innovations that it enabled.

Transportation was revolutionized with the creation of steam powered railroads cutting travel times by as much as 90% and forever changing how goods, raw materials, and people moved across the country.

Steam powered industrialized factories led to the mass production of steel which in turn allowed us to build bigger, taller buildings; larger bridges, more powerful ships, and even more efficient factories.

This period also birthed the modern oil industry, which while previously plentiful, went from a relatively low value byproduct to becoming the “liquid gold” that served as the primary fuel for a rapidly increasing industrialized world.

A few decades after Watt, the internal combustion engine gave rise to automobiles, which in turn not only drove demand for oil but over time led to the creation of suburbs, shopping malls, and a new American way of life.

So while most are aware of the importance of the steam engine, it was ultimately the railroads, steel, oil, and cars that ended up not only changing the world but creating some of the largest fortunes in human history.

We are now at the beginning of what appears to be a tectonic shift of similar proportions, an information revolution even bigger than computers or the internet before it. And just as before, the transformational potential of AI is spilling over into other sectors, like computer chips, energy production, and now notably real estate.

Exponential growth leads to unprecedented new investment

The growth of AI is occurring at a speed and scale rarely if ever seen before. ChatGPT reached an estimated 100 million monthly active users just two months after launch (arguably the fastest growing consumer application in history), while Anthropic went from $0 in revenue to nearly $50 billion in less than five years.

While the appetite from consumers may be growing at exponential rates, as with most software, there are ultimately very real physical constraints that create limitations on that scaling.

To refrain from going into full detail (which would arguably take many more pages of explanation than we can capture here) we feel it’s a fair summarization to say that the backbone of AI today is “compute” meaning the computer processing power, memory, and specialized hardware such as GPU chips that are used to train and run large language models like Claude, ChatGPT, and Gemini.

Without compute there can be no AI, and as the AI models get larger, more complex, and more people continue to use them more often…we will need more and more compute.

This reality, that compute is the limiting factor on AI, has kicked off a cycle of new investment that looks less like a once-every-decade software upgrade and more like a once-in-a-generation infrastructure buildout with tech giants such as Amazon, Microsoft, Alphabet, and Meta collectively planning to spend up to roughly $725 billion on capital expenditures in 2026 alone.

And it’s not hard to understand why. The major hyperscalers do not view AI infrastructure as a discretionary investment but increasingly as fundamental to their competitive position, with the failure to keep pace being an existential threat to their very existence.

But compute does not simply exist in the ether, it lives as a remarkably complex network of server racks full of tens of thousands of state of the art chips all housed within the most technologically advanced data centers ever created.

Where AI and real estate meet: a new kind of data center

If compute is the backbone of AI and you believe that the demand for AI will continue to scale exponentially, then the supply of new AI enabled data centers must (barring other breakthroughs in efficiency) inevitably grow at a rate to match this pace.

This fact has led McKinsey to estimate that global spending on data centers could reach $7 trillion by 2030, roughly equal to the GDP of Italy and the United Kingdom combined!

But unlike AI software, which carries little marginal cost, physical real estate properties like data centers are much more difficult to scale.

To understand this challenge, it’s useful to understand better what a typical AI data center actually looks like. These are not the data centers of decades past used to host websites or store documents in the cloud.

AI data centers are larger, requiring hundreds of acres of land and millions of square feet of building, imagine something the size of 300-500 football fields. AI workloads require significantly more power, 100 kilowatts per rack or more, which represents a roughly tenfold increase that changes the needs of the entire facility. Air cooling is no longer sufficient; liquid cooling and other novel techniques become increasingly necessary. The building itself has to be specially designed around heat, power density, redundancy, and flexibility from the beginning.

Constraints create opportunity

Though the buildings themselves may be entirely unique, designing and constructing a state of the art AI data center is still subject to the same challenges, hurdles, setbacks, and delays that are the hallmark of any large scale, complex real estate project.

The challenges begin with finding a land site capable of supporting such a massive operation, then navigating the zoning, permitting, and approval process. The design requirements are unique, requiring specialists in the field. Construction itself is a multi-billion dollar undertaking that requires a huge amount of labor and machinery, comparable to constructing multiple professional sports stadiums simultaneously. There’s a huge volume of specialized materials that must be procured, many of which have order lead times that are measured in years, not months.

And the biggest constraint of all, power.

The International Energy Agency projects global data center electricity consumption will double from approximately 415 terawatt-hours in 2024 to around 945 terawatt-hours by 2030. In the United States, that level of consumption would mean the amount of power needed for just data processing alone will exceed the total amount of electricity consumed to manufacture all energy-intensive goods today combined: aluminum, steel, cement, and chemicals.

The problem with that forecast is that such an amount of power does not exist today. And merely relying on the upgrades to the existing electrical grid could take half a decade to complete and still is unlikely to be sufficient as a solution.

This reality is why you see many of the largest technology companies responding by going directly to the source, signing long-term power purchase agreements, pursuing new nuclear and renewable energy partnerships, exploring on-site generation, and reserving capacity years before it comes online.

In other words, the challenge over the next several years is not a question of will there be enough demand from consumers or will AI products deliver enough value, but instead will we have the ability to construct enough new data centers to deliver enough compute to meet even a fraction of that total demand.

And where there are constraints, there are almost always opportunities.

Positioning Fundrise investors for success

Typically, when new industries or asset classes arise there is a period where capital supply tends to lag as traditional institutions take time to process the change and get comfortable beginning to allocate their resources to an area with little historical track record.

Often during such a window, those who are able to rely on fundamental analysis, underwriting value based on a future potential as opposed to merely historical performance, have the ability to capture outsized returns.

This is a familiar pattern and one that we and the Fundrise investor base have benefited from time and time again, whether it was the urban redevelopment of the early 2010s, the move to the Sunbelt pre-and post-covid, the creation of build-to-rent as a new housing type in the early 2020s, or the great deleveraging post-2022.

However, what we are seeing today with AI and the demand pressure flowing to data center development arguably dwarfs all those opportunities combined. In just the US, data center power demand is projected to triple within five years, an investment super-cycle requiring trillions of dollars in new investment.

And we believe Fundrise is uniquely well-positioned to capture a portion of this value for our investors.

Unlike most real estate companies, we have very real tangible expertise when it comes to investing in and building with AI…we believe we see and understand the demand side better than most given our front row seat to many of the companies pushing the frontier.

And unlike most venture investors, we have lived the real world challenges of developing, owning, and operating large scale real estate assets.

There may be only a few other asset managers in the world that have similar scale and expertise across both sectors (and to no one's surprise they too have been laser-focused on this exact same intersection).

Looking ahead

As we’ve shared before, while we remain highly optimistic about the long-term potential of AI, we recognize that over the short term the potential for large swings of volatility is very real. The dot-com bubble ballooned and then burst before the internet fundamentally changed all our lives. One should not confuse near-term investment performance with long-term value potential.

Where we are today in this super-cycle remains anyone’s guess. And data center development is an area that elicits strong emotions.

We believe AI has the potential to greatly benefit all of us, but such an outcome is not guaranteed and will require intentional action on the part of both the private and public sector. New, large scale data centers are fundamental to unlocking that potential, but the negative externalities for some are real and should be well considered as development occurs.

Goldman Sachs notes that current AI-related capital spending equates to approximately 0.8% of GDP whereas during the late-1990s telecom investment cycle, spending peaked at 1.5% of GDP, suggesting we may very well still be in the early stages.

Our goal, as always, is to position our investors to be able to capture and benefit from a significant portion of this immense, potentially generationally defining, value that otherwise would all be cut off from the average investor.

We expect to continue to increasingly invest a significant portion of the real estate focused funds into this data center build-out opportunity while being cognizant of the growing risk of elevated values and look forward to continuing to share updates and further analysis along the way.