The New Space Race: AI Infrastructure

In 1956, President Eisenhower signed legislation creating the Interstate Highway System, declaring it essential to national security and economic prosperity. Today, we face a similarly defining infrastructure moment that President Trump is rightfully focusing national efforts on. The meteoric rise of AI is bringing an insatiable demand for data centers with it, and those data farms will require an unprecedented amount of power generation and AI infrastructure. For the economy of the future, AI infrastructure is becoming as essential as power grids, highways and water systems, and the US cannot afford to fail in its buildout. With the proper coordination among the public and private sectors, the US will accomplish a historical feat and build out the necessary AI infrastructure to win the AI arms race.

Over the last decade, AI has gone from an academic pursuit to a revolutionary force that has vast implications across all sectors. While breakthroughs with popular large language models such as ChatGPT have shocked many people, the progress that is yet to come borders on the line of science fiction. Artificial general intelligence, expected to emerge within roughly 5 years, marks a cognitive revolution where AI can perform any intellectual tasks at a comparable or superior level to humans.2 Artificial super intelligence marks a world where AI’s intellectual abilities exceed the collective intelligence of humanity and is expected to be accomplished in the next 10-30 years. Artificial super intelligence would be able to recursively improve itself, learn at an incomprehensible pace, and potentially lead to technological breakthroughs not yet imagined. Artificial super intelligence marks the grand prize in the race, and whichever country and company claims the grand prize will gain an extreme amount of power with it.

With the stakes so high, both the public and private sectors in the US are keen on winning the AI competition. The White House recently released “Winning the AI Race: America’s AI Action Plan”, a plan to export American AI, promote the rapid buildout of AI infrastructure, and enable innovation in accordance with its effort to win the global AI competition.3 President Trump has likened the competition to the space race decades ago, rightfully prioritizing AI as a cornerstone in American innovation that will have broad implications for the global economy and balance of power in the world.

In the private sector, big tech has been leading by spending lavishly in its quest to win the AI race. Capital expenditures driven by AI are still going up as Meta recently upped their projections to $66-72B for the year, Google raised their guidance to $85B for the year, and Microsoft bumped their current quarter projections up to $30B.4 However, for the US to win the race, AI infrastructure and the energy that powers AI will be critical to big tech spending and government motivation.

With the demand for AI, the scale of the data centers and power behind the technology is growing exponentially. AI data centers typically require far more energy than traditional data farms, and Deloitte estimates that the power demands for these facilities will grow over 30 times in the next 10 years.6

The leading AI infrastructure developers who have been building the data facilities behind the AI boom are called hyperscalers. The biggest hyperscalers’ largest data farms currently need roughly 500 megawatts of power or less, but the largest projects that they are building will require up to 2 gigawatts of power, 4 times today’s biggest online projects.6

Further, there are 50,000 acre data farms coming down the pipeline from other hyperscalers, which could consume up to 5 gigawatts of power, ten times the power demands of today’s biggest facilities, and enough power needed for roughly five million homes.6

Power represents the biggest challenge to building out the country’s AI infrastructure backbone and will represent a major hurdle on the way to winning the AI race. A survey of 120 US-based power companies pointed to power capacity as the top challenge for AI infrastructure buildout.6 Some AI facilities are currently on a 7-year waitlist for connection to the grid and are demanding exorbitantly more power today, while the overall uncertainty of what the future holds represents a major risk of overbuilding power capacity in the future. As such, one of the other top concerns of data farm and power company executives alike was the mismatch between data farm and grid build-out timelines. While data farms only take 1-2 years to build in many cases, new power capacity can take 5 years or more to bring online.

The governmental permitting process feeds directly into timeline mismatches and is another top concern of executives. Governmental permitting timelines can take up to a year in some states and environmental impact statements alone can take 2 years to complete.6

To overcome the challenges to build the AI infrastructure necessary to win the AI competition, executives agree that technological innovation and regulatory changes are key. In other words, a collaborative effort between the public and private sector.

First, technological advances led by big tech are needed to make data center and power operations more efficient. Advances in data farm cooling, chip power delivery, and grid transmission will all be critical to lessen the energy demands of these facilities and shrink the gap between the amount of power we have today and what will be needed.

Second, regulatory change is imperative to add new power capacity. Federal and state regulatory agencies will need to cooperate to expedite permitting and environmental timelines so that new capacity can be brought onto the grid on a closer timetable to data farm needs, which will lower uncertainty around over or underbuilding capacity. Further, easing restrictions around grid interconnectivity will allow power to flow more freely throughout the country and meet peak demand. Markets with surplus power could connect to constrained grids within a year, compared to the five-year timeline required to build new power capacity. On a federal level, the Trump administration has committed to deregulation and incentives to build out data centers but could do more by promoting renewable energy, including solar, which can be deployed to the grid most quickly.

Third, data center and power companies will need to coordinate closely to budget for power needs of the future. Through partnerships, power needs will be communicated as they change and the risk of over or underbuilding new capacity on the grid can be lessened. Partnerships can also provide faster connections to the grid if data farms are willing to flexibly schedule computing tasks and commit to drawing less power during peak demand while more capacity is brought online.

The AI competition is about more than big tech’s next product. Whoever wins and achieves artificial super intelligence will wield a technological power never seen before. The proliferation of AI is driving an exponential demand for new data facilities and subsequently their power needs, which brings a unique set of challenges with it. With power being the biggest constraint holding back the data farms behind AI, an all-out sprint will be needed to build the necessary energy infrastructure to win the AI infrastructure race. Just as the US met the moment in 1956 and built out the interstate highway system or won the space race during the cold war, the US can win the AI race with the proper public and private partnerships. By incentivizing and developing technological advancements, easing federal and state regulations, incentivizing all types of power being brought online, and keeping close coordination among data centers, power companies, and regulatory agencies, the US is positioned to meet the moment once again. Through public and private partnerships, the US will build the AI infrastructure backbone for the technologies of the future and win the AI race.

Sources:

1. Wall Street Journal. July 2025. "Trump Administration Pledges to Stimulate AI Use and Exports"
2. The Science of Machine Learning & AI. August 2025. “Artificial Superintelligence”
3. The White House. July 2025. “White House Unveils America’s AI Action Plan”
4. The Wall Street Journal. July 2025. “Why Microsoft and Meta Are Soaring After Earnings”
5. The Wall Street Journal. July 2025. “Big Tech’s $400 Billion AI Spending Spree Just Got Wall Street’s Blessing”
6. Deloitte. June 2025. “Can US infrastructure keep up with the AI economy?”