• AI leaders are preparing to take America into what Sam Altman calls the “Intelligence Age.”
• Getting there will depend on building vast amounts of new AI infrastructure on US soil.
• Whether investments in this infrastructure will ever pay off is another matter.

America is ready to reach a new age of intelligence. Getting there —and staying ahead of rival nations in the AI race — depends on a plan to transform the physical world that’s becoming more formidable by the day.

Leaders driving the AI boom entered 2025 by getting louder about the radical transformation they say is needed on US soil to deliver an era of AI-led superintelligence: more data centers, more chip plants, and more power infrastructure.

By taking root in the physical world — huge data center facilities depend on complex wiring, hardware, and integration with power infrastructure across vast amounts of landmass — the hope is that AI software could one day transform society the way the Industrial Age did.

Sam Altman, the CEO of OpenAI, calls this next leap the “Intelligence Age.” In a September blog post, Altman said its defining characteristic would be “massive prosperity.” However, he cautioned that without enough infrastructure, “AI will be a very limited resource that wars get fought over and that becomes mostly a tool for rich people.”

Last week, in one of his final executive orders, President Joe Biden signaled intent to build more at home with plans to lease acres of federal land to private sector firms with the know-how to develop complex AI infrastructure. The intent to build is likely to continue after Donald Trump’s inauguration on Monday, as tech leaders rally around the incoming president and put AI among the top priorities on his agenda.

Biden’s executive order followed the release of a blueprint from OpenAI a day earlier, which claimed “the economic opportunity AI presents is too compelling to forfeit” by not building the infrastructure needed.

Data centers, power plants, and chip manufacturing plants will all cost money — a lot of money. Goldman Sachs estimates that roughly $1 trillion will be spent in the next few years alone to develop the infrastructure needed to bring today’s AI models closer to superintelligence.

It’s why the big question investors and companies must now grapple with is whether or not they are willing to put up money for a vision of the future that is hardly guaranteed.

The case for building AI infrastructure

Altman has offered no shortage of reasons for spending so much money on achieving superintelligence.

Ensuring technological hegemony over China is one. As his company said last week, “there’s an estimated $175 billion sitting in global funds awaiting investment in AI projects” that “will flow to China-backed projects” and strengthen Beijing if not directed to the US.

Another is that superintelligence could unlock unimaginable prosperity for society. Altman recently said that “if we could fast-forward a hundred years,” the prosperity from superintelligence would feel just as unimaginable as today’s world would to a lamplighter, a person employed to light and maintain street lights until about the 1950s.

The third reason is perhaps more surprising. In a blog published at the start of the year, Altman said his company is now confident that it knows how to build artificial general intelligence, a term often interchanged with superintelligence despite their differences.

It’s a combination of factors that will, in some way, have triggered the flood of comments from those who want to play their part in developing the infrastructure needed to deliver superintelligence.

In a blog published this month titled “The Golden Opportunity for American AI,” Microsoft president Brad Smith said the company planned to spend $80 billion alone this year on data centers. “Not since the invention of electricity has the United States had the opportunity it has today to harness new technology to invigorate the nation’s economy,” he said.

Last year, in conjunction with BlackRock and others, the tech giant unveiled a fund focused on AI infrastructure with an investment potential of up to $100 billion.

In an interview with Semafor last month, Google CEO Sundar Pichai said that he was ready to work on a “Manhattan Project” for AI once Donald Trump takes office, underscoring the scale of the development and investment needed by invoking the World War II program that eventually produced the atomic bomb.

Meanwhile, Japanese conglomerate SoftBank committed $100 billion to investing in the US over the next four years, focusing on AI and related infrastructure.

A risky investment

While there is clear intent to develop AI infrastructure, it’s not clear if or when the investments will pay off — for two key reasons.

First, much of the infrastructure needed in the US faces an uphill struggle to get built.

Take chip plants. US companies like Nvidia, Google, AMD, and others that specialize in designing chips have developed a significant reliance on Taiwanese firm TSMC to manufacture those chips in the Far East, where a combination of cheap, skilled labor, economies of scale, and a long history of government support for the semiconductor sector has made the incredibly expensive business of manufacturing chips easier to pull-off.

Simply throwing capital at projects aimed at getting chips manufactured in the US won’t cut it. Efforts to build chip manufacturing plants at home have been taking shape — the Biden administration’s CHIPS Act has provided billions of dollars of grants to semiconductor firms in the US — but there remains a huge gap between the capabilities of manufacturers in the East versus those at home.

The AI boom has been kind to TSMC, with its value roughly doubling last year to $1.1 trillion. US chip manufacturer Intel, meanwhile, more than halved to around $85 billion.

Clean power infrastructure, increasingly focused on nuclear power, also faces challenges. Returns on investment in nuclear power projects meant to provide clean energy to intensive data centers are highly uncertain. These projects also face significant regulatory hurdles.

In December, for instance, the States of Texas, Utah, and Washington D.C.-based company Last Energy sued the Nuclear Regulatory Commission over claims that the government agency was applying the same risk analysis to small modular reactors as it was to large-scale power plants. These SMRs, as they’re known, are meant to make access to nuclear power cheaper, given their compactness and greater affordability versus traditional nuclear plants. But even these face roadblocks.

The second big reason that investors may want to approach infrastructure investment with caution is that the emergence of superintelligence remains highly speculative.

Altman’s claim that there is now a clear path to AGI is worth taking seriously, as new models like OpenAI’s o3 released in December demonstrate increasingly sophisticated reasoning capabilities that do more than just parrot their training data.

That said, there have been rumblings across the industry recently about AI models hitting a wall in terms of performance improvements.

Without really serious advances in capabilities, then, or a clearly defined path forward to superintelligence, it is not clear how or when these colossal bets on AI infrastructure will pay off. But with China and other nations showing no sign of slowing down, it is clear that the cost of not being in the AI race could be far greater.