Inside The AI Infrastructure Buildout

This voice experience is generated by AI. Learn more.This voice experience is generated by AI. Learn more.Madhurima Manchala, Founder of Maps.ai and Early Stage Venture Capital Investor.

gettyAsk an AI model to plan a three-day trip to Tokyo, and an itinerary appears in seconds. At the center of that seemingly effortless interaction is a data center, where chips, power, cooling and networks come together to produce compute. At global scale, these facilities form the backbone of what Nvidia CEO Jensen Huang has called “the largest infrastructure buildout in human history.”​

The United States is currently leading that infrastructure buildout. More than 1,500 additional data centers are planned or under development across the country, while the world’s largest technology companies are expected to collectively spend roughly $725 billion on capital expenditures this year, with much of it going toward AI infrastructure and data centers in the United States.​

And the facilities themselves are becoming dramatically larger. A decade ago, a large data center required roughly 30 megawatts of power. Today, AI campuses commonly reach 200 megawatts, with some reaching gigawatt scale. That growth is colliding with physical constraints: land, skilled labor, electrical equipment, grid connections, power and cooling are all becoming harder to secure.

Within these constraints, power may be the most consequential. The United States may lead the data center race today, but maintaining that advantage could increasingly depend on whether its physical and energy infrastructure can expand quickly enough to support the next wave of compute. The buildout is also meeting resistance from local communities and policymakers​ as concerns over power, water and land use prompt restrictions and moratoriums on new data center development.

I believe these factors collectively form the central tension of the AI infrastructure race. The pressure to scale is creating opportunities for innovation across the physical data center stack, from construction and power generation to cooling, operations, maintenance and ownership. Having worked at the intersection of venture capital and the built world, I wanted to better understand where innovation in this sector is happening and what it unlocks. This article explores what I’ve learned about the technologies that could allow data center infrastructure to scale rapidly and more responsibly.

The data center boom is beginning to change the construction industry itself. According to Statista, “Private sector spending on data center construction in the United States reached $41.1 billion in 2025, up from $1.8 billion in 2014.”

But demand is growing faster than the industry’s ability to build. Electricians and other skilled trades are particularly scarce. However, the labor shortage is also becoming a catalyst for innovation, leading to new ways to find, train and deploy skilled workers while automating some of the most repetitive and dangerous work on the jobsite. Skillit, for example, uses AI to connect contractors with vetted construction workers; robotic systems such as Rugged Robotics can read digital construction drawings and autonomously mark layouts across data hall floors; and other machines are being tested for physically demanding tasks such as tying rebar and drilling overhead.

Many investors are betting that automation can extend to heavier work, as well. Construction robotics companies focused on autonomous equipment have recently attracted significant funding, with 10 robotics and semiconductor-related startups joining the unicorn club as of February 2026. But the emerging model I am seeing is less about replacing workers than allowing scarce skilled labor to be deployed more effectively. I expect we will increasingly see a world where robots and humans work alongside each other rather than one replacing the other.

Technology is also being applied before workers reach the jobsite. Contractors are experimenting with augmented and virtual reality to shorten training cycles and transfer field knowledge to new tradespeople more quickly and safely. At the same time, more work is moving off-site. I have observed an increase in electrical, mechanical and structural systems that can be prefabricated in controlled environments and delivered as ready-to-install modules—an approach that seems particularly well-suited to the repeatable design of data centers.

Together, these technologies are helping the industry build faster. But even as construction becomes more efficient, it runs into a harder physical constraint: power. Every new data center still needs enormous amounts of electricity, making energy availability increasingly central to where and how quickly AI infrastructure can grow​.

Building faster only matters if there is enough electricity to turn the machines on. In parts of the United States, connecting a new data center to the grid can take more than a decade. That constraint is driving energy innovation. Developers are exploring behind-the-meter generation, batteries and new sources of firm power. For example, Form Energy recently raised $750 million​ to expand its iron-air batteries, which can provide power for up to 100 hours. And advanced geothermal and small modular nuclear reactors are attracting interest as potential sources of round-the-clock clean electricity. The IEA estimates that data centers could deploy 20 to 25 GW of battery storage globally by 2030.

More electricity, however, means more heat. As AI racks grow denser, traditional air cooling is giving way to direct-to-chip liquid cooling, rear-door heat exchangers and closed-loop systems that recirculate water. Related Digital, for instance, has agreed to refrain from using “high-water-use evaporative cooling“ (download required) to allay local concerns about high water consumption.

Innovation continues after the facility opens. Companies such as Thalo Labs are combining sensors with AI to monitor HVAC equipment and identify problems before they become costly failures, shifting maintenance from reactive to predictive.

The data center of the future may look very different from today’s sprawling campuses. Compute could become more modular, distributed and self-sufficient, moving closer to abundant energy or even underwater and into orbit. I believe the next infrastructure race will be about more than scale: It will be about reinventing where and how computing happens.

The information provided here is not investment, tax or financial advice. You should consult with a licensed professional for advice concerning your specific situation.

Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?