Talk of the “AI race” is everywhere these days.
The competition to master this technology shapes everything from high-level tensions between the US and China to the local politics of where to build data centers.
And the stakes are high. Whoever masters AI will “rule the world,” according to Russian President Vladimir Putin. US President Donald Trump has been blunter: “Whoever wins AI, wins.”
But what exactly is that “race”? Who’s competing? What’s the prize? And, importantly, how do we know who’s winning?
To demystify it all, GZERO’s Alex Kliment sat down with Heather West, who leads Eurasia Group’s Geotechnology practice. Their conversation has been edited for length and clarity. Also, it features an informed cameo from Heather’s dog.
Alex: Heather, everyone talks about “the AI race.” What is it?
Heather: People talk about it as if it’s one thing, but there are really a lot of different races that people are running at once.
The United States wants to win the most capable model race, and they’re winning it.
China really wants to win the diffusion race to get more people around the world to use their models — and they’re winning it.
Europe wants to win the regulation race, making sure there are rules that protect rights around AI and technology — and they’re winning it.
How do we know who’s “winning” these races?
Well, everyone’s measuring differently. The frontier race, for example, is about the most capable general-use AI models, as evaluated both by benchmarking organizations and by market adoption. And, according to most measures, American models get the top marks.
Why is the US so clearly in the lead there?
The top reason is all the technological research and development talent that has been moving to Silicon Valley for decades. Second is the sheer amount of investment in the United States. You don’t see that in other countries — and that’s partly because access to capital in the United States is so much greater, but also because there’s a very risk-tolerant venture community which is excited about these investments.
And why is China winning the diffusion race?
Chinese models tend to be cheaper and largely as effective for most tasks. Also, some people are worried about relying on US models that get powerful enough that the government suddenly puts export controls in place again.
So that’s another arg–
Sorry, my dog’s jumping all over me and I have no idea why.
Because he’s excited about the AI race, that’s why.
Obviously! So, as I was saying, the perception of more reliable access is another argument for using the self-hosted Chinese open-weight models instead of the US ones.
Meanwhile, on racetrack #3, you said Europe is winning the regulatory race. How?
Well they don’t have the same research and capital and talent base [as the US and China], but they are the model that many people point to when it comes to regulation.
The EU was the first to put out comprehensive AI regulation a few years back, with the AI Act. That was before “generative AI” was really a household term, so some think of it as a little bit out of date, but they haven’t actually finalized the implementation yet.
Still, the responsibilities for someone deploying AI in Europe are much clearer than in many other jurisdictions.
Interesting. So, what’s the prize for winning any — or all — of these races?
World domination?
I like it.
I seriously think that’s partially the theory, but depending on which race you’re running, it’s a little bit different. If you’re in the EU, your theory is that clear responsibilities create room to run for companies. In the US, it’s to stay at the frontier at all costs. This has the AI labs and the US government with their foot on the gas.
And if you’re China, it’s that although your models continue to be four to eight months behind the US frontier, they will be cheaper, easier to control, and have more guaranteed access than the US models — even if the US models are more advanced.
It sounds like the AI race is so far about the US, China, and the EU. Is there anyone else who could hop onto the track?
Absolutely. There are credible middle power coalitions that could come together — not to challenge dominance in the whole AI stack, but to challenge pieces of it.
So far we’ve been talking about the AI race as if it’s only dependent on AI models, but AI models are constrained by compute [i.e., the computing power needed to run them — ed.], and that’s constrained by data centers, which need power, water, connectivity, and big open spaces. There are many places that could actually become compute superpowers.
Folks talk about Canada, they have wide open spaces, they have good natural resources, they have good access to power, and they have a really smart workforce. So there have been conversations about the EU and Canada teaming up, and that actually is a pretty credible idea.
The Gulf, meanwhile, has enough capital and currently has access to chips that they could be a credible contender too.
Lots of players. Let’s end with this: if you could win just one of the three races yourself, which one would you want to win?
If I chose one and I knew I would win?
Yeah.
I think diffusion.
Why?
If I throw a party and no one shows up, what’s the point?
