A Race for the World

The illustration above was created for this article via prompts in OpenAI’s image creation tool DALL-E.

Before making his name working in Washington and Silicon Valley, Chris Lehane was a history major at Amherst. In his telling of it, technology took the world from the Stone Age to the Agricultural Age and then to the Industrial Age. Now he thinks artificial intelligence has us poised to enter an entirely new era, the Intelligence Age. 

Lehane is the Chief Global Affairs Officer at OpenAI, one of the most critical roles at arguably the most consequential company shaping artificial intelligence. OpenAI’s launch of ChatGPT in late 2022 was the starting gun that set the AI race into motion—and as that race intensifies, so does the pressure Lehane feels.

“AI will shape the future of US national security and economic competitiveness, and transform everything from education to healthcare, so the stakes couldn’t be higher,” he said. “I wouldn’t typically quote Vladimir Putin, but he said whoever wins this contest wins the world. And I think he’s right.”

Washington—which Lehane understands better than most in Silicon Valley—is waking up to the importance of the AI race. In mid-December, the bipartisan House AI Task Force published a report with 89 separate policy recommendations. “I’ve spent enough time in Washington to know how hard it is to reach bipartisan agreement on any major issue, especially one as complex as AI,” Lehane wrote recently. “The fact that this report drew the unanimous support of lawmakers from both parties is a reminder that AI transcends politics.”

In a discussion with Brunswick Partner Ash Spiegelberg, who co-leads the firm’s tech, media and telecommunications group, Lehane spoke about the need for bold action on AI.

The days between their conversation taking place and being published offered a glimpse of how quickly and profoundly the AI landscape can change. In that time, President Trump announced a $500 billion Stargate Project to build new AI infrastructure for OpenAI in the US—a project whose funders include SoftBank, OpenAI and Oracle, and whose technology partners include Arm and Nvidia.

The second development proved Lehane’s prescience: a new AI model launched by DeepSeek, a Chinese tech startup, sent shockwaves not only for its sophistication but also because it appeared to require less expensive chips and fewer resources. Legendary tech investor Marc Andreessen called the model’s launch “AI’s Sputnik moment”—which Lehane had warned Spiegelberg was coming.

Where are we right now with AI? And what might we expect to see in the not-too-distant future?

When Steve Jobs first explained what a computer was, he talked about it being a bicycle for the brain. What OpenAI released in November 2022 with ChatGPT—which is really ChatGPT 2.5—was like a 10-speed bike for the brain. What we released in September—our o1 model, which has the capacity to do reasoning—you could consider an electric bike. What’s coming in the near future will be a motorcycle for the brain. And after that, maybe a fleet of motorcycles. Eventually, a flying fleet of motorcycles.

Another way to think about it is on the scale of electricity. Electricity transformed how we live, work, connect with each other—even how we play. Sam Altman, our CEO and co-founder, talks about AI leading us into a new Intelligence Age. It’s going to create enormous growth and, with it, incredible opportunities to share that prosperity.

If you zoom out and think about what AI is, it’s giving people tools to help solve really hard problems. But if you want to have an AI that benefits everyone, where people are able to use this as a tool for themselves, then we need to build democratic AI that aligns with core values like individual choice, privacy, respect for the rule of law and a commitment to making the technology accessible to all rather than being centralized in the hands of a powerful few.

You’ve said that the world faces a choice: democratic AI or autocratic AI. Is it really that simple and that stark?

Right now, objectively, there are two countries that can build AI at scale: the United States and the People’s Republic of China. This is a zero-sum game: there’s about $175 billion in global infrastructure funding waiting to be invested, and the only question is whether it ultimately goes to US-backed projects or China-backed ones. I feel very strongly that the US is going to win. Our democratic system, and the democratic values built into it, are inherent superpowers. But right now, the US has somewhere between a six-month and three-year advantage.

How do you measure each country’s relative strengths in AI?

We think about this in terms of a “compute gap.” The key elements are chips, data and energy—if you put them together and marshal them, they give you compute. It’s compute that allows you to build AI and keep innovating it.

If you were imagining a chart of this competition, the US is still ahead, but China is accelerating at a faster rate. Where does the PRC have advantages? On the data side, because they can scoop up as much data as they want. They have enormous amounts of energy—they’ve built more nuclear power in the past 10 years than the US has in the last 40.

I’m not a computer scientist or a mathematician or a physicist, so I default to historical analogies as a history major. Sputnik was a major wake-up moment for the US in the ’50s; its launch surprised a lot of folks with how advanced the then-Soviet Union was technologically.

We could be heading towards a Sputnik moment for AI, too, because China is working really hard to improve its models. It’s really important for people to understand that this is not an abstract concept. This is an incredibly competitive race with enormous consequences.

The US is ahead right now because of our innovators, entrepreneurs, builders and thinkers—an incredible pool of talent. But to stay ahead, we need to ensure they have access to the compute power they need to keep pushing forward.

You say “infrastructure is destiny.” Yet there’s a perception that democracies have an infrastructure problem—they tend to be slower in planning, approving and building large-scale projects. Can the US overcome this?

Infrastructure is destiny—no infrastructure, no compute. So maintaining our lead comes down to: Can we create a smart, public-private approach to infrastructure?

In the US, big breakthroughs have always involved collaboration between government and the private sector. Take the auto industry—it flourished because the federal and state governments worked together to unleash American innovation. The interstate highway system under Eisenhower was another massive build-out, done for both security and commerce.

Then there’s the internet. Back in 1996, I was in the White House when Democrats, Republicans, President Clinton and Vice President Gore all recognized that the World Wide Web—that’s what we called it then—was important. The initial instinct was to regulate it like technology that already existed, TV or radio, but those laws didn’t fit. Instead, they focused on ensuring this innovation would be American-led and built by American talent. So you had the 1996 Telecommunications Act. Fast forward to today: Most of the major digital companies in the world are US-based.

That’s what it’ll take: thinking big, acting big and building big.

The good news is that the US has a history of doing that. And today, national security remains an issue that tends to galvanize Democrats and Republicans and produce political action.

National security is also one of the rare areas where the public and private sectors have historically worked together—and one where we believe they can, and should, work together again. We just released a new Economic Blueprint that outlines the specific ways that companies like OpenAI can work with the government to extend America’s global leadership in AI innovation, ensure equitable access to AI and drive economic growth across the country.

When it comes to national security, that would include things like developing export policies that would allow the US to share frontier models with allies and partners while ensuring those cutting-edge models don’t wind up in the hands of countries that don’t share our values or commitment to personal freedom, economic opportunity and the rule of law.

OpenAI also recently released its Infrastructure Blueprint. That calls for creating “AI economic zones” in the US. What does that look like, exactly?

Since the beginning of what we would call the Digital Age, say the mid-90s to today, a disproportionate amount of the benefits have flowed to the coasts from that economic expansion, growth and development.

With AI, because it is going to be an infrastructure-type technology—like electricity—how do we think about those benefits reaching the entire country? Take Kansas. It’s a place where people are building data centers. Through a blend of natural gas, wind and solar energy, Kansas can handle the energy component, which is crucial. But Kansas also has a ton of data on agriculture—maybe as much data as any place in the world. Now imagine building a one-gigawatt facility in Kansas, where you allocate 250 megawatts to the university system there to specifically stand up an AI agriculture model that would not just serve Kansas, but the entire country—and maybe even the world. Then you attach visas to that project, so you attract high-skilled workers from all over the world to come to Kansas. You can start to see how an entire AI ecosystem takes shape, creating enormous opportunities for people who are from Kansas, born in Kansas and want to stay in Kansas.

A lot of these great states have effectively been greenhouses for the two coasts—these places grow incredible talent, and the talent goes off to the coasts. I say this as a kid from Maine, which is on the coast but has a similar dynamic. If you think about AI as infrastructure, then it has both a reindustrialization aspect for the US, and the ability to reinvigorate the American Dream.

I’ve got to ask you about nuclear submarines—that’s not a topic that comes up a lot when people talk about AI.

Without energy, none of this works. The United States Navy operates around 100 small modular reactors (SMRs) on its fleet of nuclear submarines. These reactors are roughly 250 to 300 megawatts each. If we can put small nuclear
reactors on submarines that travel all over the world, sometimes in pretty volatile places, then we should absolutely be able to build those reactors safely here on stable ground in the US.

Should Washington lead on AI policy, or is there a role for states as well?

It’s not just that there’s a role for states to play; it’s that the US will only keep its lead in the AI race if both Washington and the states help set the rules of the road for the technology.

To be more specific, you could see a scenario where Washington focuses on setting policy for the aspects of AI that impact the entire country, like ensuring the safety of frontier AI systems because of their implications for US national security and economic competitiveness. States could craft targeted policies to protect children, prevent the proliferation of deepfakes, help users know if they’re seeing AI-generated content and address other areas where AI impacts daily life.

We’ve gotten this balance right before. Michigan emerged as the center of the US automobile industry because policymakers in Lansing and Washington passed state and federal laws which helped create thriving ecosystems of automotive manufacturers and suppliers, provided federal funding for highway construction projects and offered support for workforce training programs. Generations of workers were able to enjoy their fair share of the American Dream, and Detroit and its surrounding communities became part of the Arsenal of Democracy that helped to win World War II.

Another place where AI is in focus—particularly when it comes to regulation—is the EU.

A lot of people look at AI as if it’s the next version of social media. And I think many policymakers and businesses in Europe felt social media was an extractive technology. Yes, it helped people connect and engage, but the sense was these companies were able to access data, monetize that data and the flow of capital went from data in Europe to monetization outside of Europe.

Folks have to understand that AI is not social media. It’s not an extractive technology, it’s actually a productive technology—if it is built, developed, used, deployed and accessible in these countries. That’s incredibly important for Europe to understand, because there’s a limited window for countries to position themselves and make those necessary investments to get those productive benefits.

Look at the Draghi Report, which found that the EU is grappling with slow growth, aging populations and innovation gaps. Germany and France have energy. They have access to rich, unique data. They have talent. How do you take advantage of that?

You’re British, Ash, so I’ll use an example from the UK—think of the National Health Service. It’s probably sitting on top of as much healthcare data as any entity in the world. The UK has some of the leading talent already in the world. They have the gold of the AI era—data and talent. If the UK created the leading healthcare model in the world, not only would that serve NHS, but it would also obviously help people in the UK—they’d enjoy better care, and maybe solve diseases we’d never thought were solvable.

But even more, that model could become a global, commercial entity for the UK, which both has soft power components to it and real direct economic benefits.

I know these things are easy to talk through and talk about, but let’s not lose sight of the fact that these are genuine possibilities. The question is whether you act boldly enough to take advantage of them.