Forget Data Centers. This Asset Class Is the Next Big Infrastructure Bet.

There’s a new asset class on the block. It barely existed a few years ago and it’s growing like crazy. It requires tons of energy, dedicated grid connections, and significant up-front capex. It powers a technology that’s transforming the world, and there’s a global race to build it out. 

And no – I’m not talking about data centers.

It’s been 100 years since cars replaced horses in America. We built huge amounts of infrastructure to enable this shift, including 150,000 gas stations and 190,000 miles of pipeline. It transformed our economy and our environment. New business models emerged, from motels and drive-thrus to parking garages. Suburbs and freeways changed how most of us lived. 

The next shift is already underway. In 2009, four of the top ten global companies by market cap were oil and gas. Today, eight out of ten are tech. The most dynamic parts of the economy run on the grid, and investment is concentrating in sectors powered by electricity. Electrons, not molecules, are increasingly driving value creation and growth. And this new electron economy requires new kinds of infrastructure. 

Data centers are the most obvious example. Instead of steel or crude, they rely on data and power; instead of ports or railways they need internet nodes and water. And they account for a massive share of the economy. In 2017, at the peak of the shale oil buildout, total oil and gas capex hit $85 billion. Today tech companies spend that much on AI infrastructure every 2 months.

But the data center boom is just the first visible sign of a much deeper shift in the economy. If you look past the big bets on generative products, AI is starting to show up in the physical world. That’s creating whole new industries. Some of these might seem as unfamiliar as a refinery would to a carriage driver, but look closely and you can see a new generation of businesses racing ahead to meet the emerging demand.  

One example is launchpads: fleet-scale charging platforms for autonomous vehicles and high-throughput EVs. These are powered sites that deliver electricity and data ops for robotaxis and heavy-duty trucks. Like data centers they sit at the crossroads of land, power and technology – and like data centers they’re rapidly accelerating because of AI.

This time two years ago Waymo was publicly available in one city: Phoenix. Today they’re in Los Angeles, San Francisco, Austin, and Atlanta, and expanding to Miami, Dallas, Houston, San Antonio, Orlando, Las Vegas, Detroit, Denver, Nashville, Seattle, Tokyo, and London. Zoox operates in Las Vegas and is launching in San Francisco. The robotaxi fleet is set to grow 20x by 2030, with autonomous commercial trucks not far behind. 

Self-driving tech is already here. It’s safer, cleaner and getting better every day. But the one thing autonomous vehicles need is reliable access to power. In other words, in order for this technology to grow, it will need infrastructure to run on.

Fleet operators need a network of convenient locations to charge and operate their vehicles. These sites are at the intersection of physical and digital infrastructure, they’re different from other kinds of assets in a couple of important ways. 

First, location. Avoiding ‘Deadhead’ miles – trips to refuel, reload or get to waiting passengers with an empty vehicle – is mission critical for rideshare and logistics businesses. Heavy-duty trucking sites have to be close to ports, freeways and logistics hubs; rideshare hubs need to be close to population centers.  

Second, power. A site for charging fleets can need 10MW or more of power capacity. That’s less than data centers, but an order of magnitude more than most commercial buildings. Vehicles need guaranteed power when they arrive, which requires a site to provide dynamic load management, have contracts with utilities that allow demand to flex up and down, and state-of-the-art electrical infrastructure. 

Sites that meet these requirements are scarce  and take multiple years to build, along with, significant capital and deep expertise across real estate, planning, engineering, operations and technology to develop them. Fleet operators don’t have the time, experience or balance sheet to do it. In a maturing market, it’s more efficient for specialized companies to build out the network. That means not just putting up charging stations, but building a full stack service that AV companies can use to minimize operational costs. 

This won’t be easy. For one thing, new sources of demand have to be integrated with the grid. We predict EV charging demand in the US will hit 100 TWh per year by 2030.This is small compared to where data center loads will be in 2030, but getting power from the distribution network in established cities is hard, and the spread of AI into more sectors will increase the demand for electric grid capacity for many uses. Sites need to be built with future demand in mind: batteries are 5x cheaper and 50% more energy-dense than a decade ago and still improving, which means they deliver more miles, but will also require more power. Buy-in from local communities is critical, and local governments will need to develop a model for zoning them. 

Finding sites that can meet all three criteria – enough power, in the right location and with appropriate zoning – is difficult. We also need new models to finance the buildout of a network to power the mobility hyperscalers. Estimates suggest $5-7B of new investment in the US by 2030 to open new markets. That means bringing in scalable sources of both equity and debt.

The challenges are real, but so are the opportunities. Applied AI is already creating the demand for new kinds of infrastructure. As a greater share of economic activity shifts to electricity and AI shows in industries beyond tech, these will become as familiar as gas stations and powerlines are today – the infrastructure that powers our world. The future isn’t far away.