Chasing AGI sounds bold and visionary. It also sounds like the kind of goal you can use to justify almost anything—endless spending, endless hype, endless patience from the public—while the other side quietly ships products that change daily life.
That’s the uncomfortable tension in a new report from a Washington think tank that compares how the U.S. and China are approaching AI. The headline contrast is simple: American firms are pouring energy into the moonshot of artificial general intelligence and building out huge data centers to support it. Chinese companies, by comparison, are pushing for efficiency and fast integration of AI into real-world stuff: robotaxis, vehicles, wearables.
If that’s even mostly true, I don’t love what it says about our priorities.
Because “AGI first” is a nice story to tell investors and engineers. It’s clean. It’s dramatic. It flatters the people building it. And it keeps the center of attention on a small set of companies that can afford massive compute. You get to say you’re building the future, not just improving a product.
But “efficiency and integration” is how you actually win the boring parts of the world. It’s how you get AI into the car someone buys, the device they wear, the delivery system they rely on, the customer service line they call, the warehouse that fills the order. That’s not as romantic as AGI, but it’s how tech becomes infrastructure. And once it’s infrastructure, it’s very hard for outsiders to displace.
Imagine two futures.
In the first, the U.S. keeps aiming up and out. We build bigger models, bigger data centers, bigger ambitions. The demos get more impressive. The headlines get louder. Meanwhile, everyday AI remains clumsy in the places people feel it most: hospitals, schools, city services, small businesses. The “real world” stays messy and under-invested because the prestige is elsewhere.
In the second, China does what the report suggests: it treats AI less like a distant finish line and more like a tool to be embedded everywhere. A robotaxi rollout isn’t just a robotaxi rollout. It pressures regulators, trains the public, forces the supply chain to mature, and creates a feedback loop where usage creates better systems which creates more usage. Wearables aren’t just gadgets; they’re data pipelines and habit builders. Vehicles aren’t just transportation; they’re moving computers that can become platforms.
If you want to argue with me, here’s the debate: moonshots versus plumbing.
Moonshots create breakthroughs. Plumbing creates leverage.
The U.S. has always been good at moonshots. But we sometimes treat plumbing like beneath us—until we wake up and realize someone else owns the pipes. That’s when “efficiency” stops sounding like a technical detail and starts sounding like power.
There’s also a risk hiding inside our obsession with AGI: it rewards centralization. If the main path forward is “more compute,” then the winners are whoever can raise the most money, secure the most chips, and build the most data centers. That narrows the field. It turns AI into an arms race where a handful of players define what’s possible and everyone else rents access.
And renting access is fine—until it isn’t.
Say you’re a startup trying to build something genuinely useful for nurses, or truckers, or teachers. Your big problem isn’t that the model isn’t smart enough in some abstract way. Your problem is cost, reliability, and whether it works at 7:30 a.m. on a bad internet connection with a tired human on the other end. Efficiency is the difference between “cool demo” and “this is now normal.”
Or say you’re a city deciding whether to allow autonomous vehicles. If one country builds the most reliable systems and deploys them at scale, it doesn’t just sell vehicles. It sets expectations. It builds the playbook. It shapes what “safe enough” means. It gains confidence and competence that compound over time.
Now, to be fair: the U.S. focus on AGI isn’t pure ego. If you believe a step-change in capability is coming, you’d be foolish not to chase it. The upside is enormous. If AGI-level systems arrive, the country that leads there could shape standards, capture value, and maybe even decide how these systems get governed.
But that’s also the problem: “if.” Betting everything on a big “if” is not a strategy; it’s a personality trait.
We don’t actually know whether AGI is around the corner, or what it will look like, or whether it will land as a controlled tool or an unstable mess. We also don’t know whether the practical wins—robotaxis, AI in cars, wearables—will be smooth or hit walls like safety, public trust, or simple consumer boredom. The report describes a contrast in direction, but direction doesn’t guarantee arrival.
Still, I can’t shake the feeling that the “integration” path has a quieter kind of inevitability. When technology gets embedded into routines, it stops being optional. And when it stops being optional, it starts shaping labor, habits, security, and even politics.
So here’s the real stakes question: if one country focuses on the biggest possible brain and the other focuses on putting decent brains everywhere, which approach actually determines who holds power in the real world?