Infrastructure is destiny in the AI era

In a presidential campaign where the two candidates don’t agree about much of anything, there’s one area of common ground: the importance of the U.S. being a clear leader in AI.

Reports indicate Trump allies want to create a series of “Manhattan Projects”  to identify new uses for AI technology. In her acceptance speech at the Democratic convention, Vice President Kamala Harris promised to make sure that the U.S. would “lead the world into the future” on AI and that “America, not China, wins the competition for the 21st century.” 

The comments show that both candidates understand an essential truth: When it comes to the technology that will shape so much of what the world will look like going forward, infrastructure is destiny. It’s increasingly clear that building more data centers and power plants will catalyze a reindustrialization of the U.S. that benefits the entire country.  

Investing in America’s AI infrastructure will result in tens of thousands of new jobs, significant GDP growth, a modernized grid powered by renewables like wind and solar and supplemented by nuclear energy and the construction of a state-of-the-art network of semiconductor manufacturing facilities that will invigorate local economies across the country. 

Given the stakes, we need to think big — and act big.

While much of the world understandably thinks of AI in terms of systems like ChatGPT, at OpenAI, where I am head of global affairs, we’ve just released a groundbreaking model capable of using reasoning to think its way through a question rather than predicting an answer. 

This new tool brings us closer to what OpenAI CEO Sam Altman described in a new essay as "The Intelligence Age; a period where AI will spark new scientific advances and generate unprecedented levels of shared prosperity. 

As Altman laid out, technology brought the world from the Stone Age to the Agricultural Age and then to the Industrial Age. He said “the path to the Intelligence Age” requires three factors: “compute” — things that require increased computation and processing power — “energy and human will.” 

The problem is that these demands far outstrip the available supply. Left unaddressed, the AI infrastructure shortfall could slow the pace of innovation, limit access and prevent much of the world from accessing its benefits. Bringing the world into the Intelligence Age will require building more infrastructure — and fast. 

The need is clear, and time is not on our side. J.P. Morgan estimates that capital spending on AI already rivals the mainframe era of the late 1960s and the fiber optic deployment of the late 1990s — and there’s an estimated $175 billion in global infrastructure funds waiting to be invested, according to an outside estimate calculated for OpenAI earlier this year. 

The question isn’t whether that money will be spent, but where. If we don't channel it into infrastructure projects here at home that support democratic AI ecosystems around the world the funds will flow to Chinese-backed infrastructure projects that limit access to AI and enshrine government control over its own people. 

For economic and national security reasons, we can’t allow that to happen. To ensure that infrastructure investments benefit everyone, we need a strategy to maximize access to AI and secure America’s leadership role in the technology, now and into the future. 

To understand why, it’s important to recognize that generative AI models like ChatGPT don’t rely solely on algorithms and advanced code. They depend upon sprawling campuses of buildings that house chips and servers. These data centers extend over large swaths of land and require enormous amounts of electricity to power and cool the systems. 

OpenAI is evaluating potential partnerships with industry partners and private investors to raise funding to construct a new generation of AI-specific data center campuses. Each new facility would be several times larger than the biggest data centers in the U.S. today, allowing them to process unprecedented amounts of data at speeds currently unimaginable.  

As a thought exercise, OpenAI recently engaged outside experts to work with us to analyze the potential job gains and GDP growth that would come from building 5GW data centers — each of which would draw 5 billion watts of electrical power simultaneously — in various locations across the U.S. 

We found that constructing and operating a single 5 GW data center could create or support about 40,000 jobs — ranging from construction and maintenance positions to jobs in restaurants, retail and other industries that would serve the new workers — and contribute between $17 billion and $20 billion to a state’s GDP. 

In Texas, our analysis found that constructing a 5 GW data center could create and support more than 48,000 jobs and boost the state’s economy by more than $20 billion per year. West Virginia could see 32,000 jobs and $16 billion in economic gains, while Wisconsin could see 38,000 jobs and $16 billion in gains. 

Similar gains would come from building multiple 1 GW facilities in different states. Maintaining each center, regardless of its size, could mean thousands more jobs and billions in additional GDP growth. 

Additional jobs would be created as a result of any push to build and operate the new semiconductor manufacturing factories and energy production facilities, including the development of new types of energy sources, that would be needed to equip and run the centers. 

Those gains will improve the lives of countless Americans while revitalizing local economies, particularly in states that have suffered from the loss of manufacturing jobs or the decline of other industries. Achieving them will require us to go big and think creatively. 

As OpenAI has talked with thoughtful public and private sector leaders, a number of big ideas have come through.

First, local, state and federal governments could create “AI Economic Zones” that significantly accelerate the permitting process for infrastructure construction.  That would make it easier to build new solar arrays and wind farms and to bring unused nuclear reactors back online. 

Microsoft and Constellation Energy announced a deal to restart a reactor at Pennsylvania’s Three Mile Island nuclear plant, but it will take years to secure the approvals necessary to bring the reactor back online.

Next, the federal government could dramatically ramp up its spending on power and data transmission and streamline the regulatory approval process for new lines. That would be accompanied by the creation of a “National AI Infrastructure Highway” to connect those regional power and communication grids in the interest of national economic competitiveness and security.

And third, states that provide subsidies or other assistance to companies launching infrastructure projects could insist that a meaningful percentage of the new computing power be made available to public universities to help create AI hubs that are closely tied to their key industries.

Major agricultural states like Kansas could establish hubs dedicated to applying AI in farming. States with strong energy sectors like Texas and Pennsylvania could develop centers focused on integrating AI into power production and grid resilience.

There are other proposals worth considering and potentially putting into place, and I know that doing so won’t be easy. Raising funds and identifying locations for new infrastructure will take time, and building the facilities will require hiring tens of thousands of workers while navigating local, state and federal regulations. Securing enough specialized chips for AI models will pose an additional challenge. 

Still, I’m confident that it can be done. By investing in AI infrastructure at home, we can catalyze the reindustrialization of the U.S., create high-quality jobs, stimulate economic growth and reinforce our global leadership — all while upholding and promoting the democratic values that define us.

Chris Lehane OpenAI's is head of global affairs.