America Can't Afford Another Failure of Imagination

When a staff person came into my office in the Longworth House Office Building on the morning of September 11, 2001, we were already on edge. We had watched two hijacked airliners hit the World Trade Center. Then the staffer said, “Congressman, look out your window—there's smoke coming from the Pentagon.”

We scanned the television channels. For several minutes, none mentioned the Pentagon. I remember thinking: If we don't know about smoke over the Pentagon, what else don't we know?

As we approach the 25th anniversary of that day, I find myself asking a version of that question about a very different strategic challenge.

The comparison should not be misunderstood. China is a strategic competitor, not a terrorist organization, and the competition over artificial intelligence bears no resemblance to the attacks of September 11. The lesson I carry from that day is about something else: the danger of failing to imagine a consequential threat until its effects are already upon us.

After September 11, the commission formed to determine what America had missed and reached a haunting conclusion: We had suffered a “failure of imagination.”

We need imagination now to picture what Chinese AI leadership could mean: a People's Liberation Army with an AI advantage over ours; Chinese companies leading century-defining industries; and countries around the world increasingly dependent on a technology ecosystem whose standards and architecture are shaped in Beijing rather than Washington and allied capitals.

This isn't an argument for recklessly building AI. Concerns about jobs, privacy and power concentrated in algorithms beyond meaningful human control are legitimate. But those concerns aren't answered by ceding leadership to Beijing.

The case for American AI leadership isn't that the technology is risk-free. It is that those risks are more likely to be managed in a system accountable to courts, a free press and voters than in one that is not.

The strategic danger is easy to overlook because this competition won't announce its outcome with smoke over a building. There will be no single day when every network interrupts programming.

Instead, the contest will unfold through thousands of seemingly ordinary decisions: how quickly power can connect to a data center; how long infrastructure sits in permitting; whether sufficient chips and electricity are available; and which model a developer in Lagos, Jakarta or São Paulo chooses to build on.

Those seemingly ordinary choices can quietly compound into strategic advantage. By the time the consequences become obvious, reversing them may be enormously difficult.

Winning this quieter contest therefore requires building.

A frontier AI model can improve in months. The power plants, transmission lines and data centers needed to deploy it can take years to approve and construct. China has spent years steering computing infrastructure toward power-rich regions through its “East Data, West Computing” initiative. In America, more than 2,000 gigawatts of proposed generation and storage were waiting to connect to the electrical grid at the end of 2025, with projects that succeed taking longer on average to complete the process.

Data centers are more than buildings filled with servers. They are where the AI race becomes physical—where advanced chips, electricity, networks, capital and algorithms come together. If America leads in models but cannot build the infrastructure necessary to deploy them at scale, technological leadership won't necessarily translate into economic or military advantage.

Communities asked to host this infrastructure nevertheless have legitimate questions about electricity rates, water use and development. National security doesn't erase them.

The answer is a fair compact, not a mandate. Developers, rather than existing ratepayers, should bear the infrastructure costs their projects create. Washington and the states should recognize the need for more electrical generation and respond with urgency. Projects meeting clear standards should receive timely, predictable permitting decisions rather than years of limbo.

As the 25th anniversary approaches, I mourn all those we lost on September 11 and in the long struggle that followed. The lesson I draw from that experience isn't that today's strategic competition resembles that terrible day. It is that America cannot afford to wait for a crisis before recognizing what is at stake.

This time, we should imagine the consequences before they arrive.

Mark R. Kennedy is founding director of NYU’s Wahba Initiative for Strategic Competition. He served in the U.S. House of Representatives from 2001 to 2007.