Last week a Chinese lab called Moonshot put a model named Kimi K3 up for anyone to download. Open weights. Roughly at parity with the best American models on the things that matter, at a fraction of the price per token. The American reflex was the one we always reach for now: panic about China.
Wrong reflex. At BNS, we build the power and networks this technology actually runs on, so let me tell you what I see, and it is not a story about them. It is a story about us.
Here is the thing almost nobody in the noise has said plainly. Intelligence is becoming a commodity, and its price is set in kilowatts. Once the models are close enough and cheap enough, the advantage does not go to whoever has the cleverest algorithm. It goes to whoever can turn power into deployed compute the fastest. That is not a chip problem or a model problem. It is a power problem, a permitting problem, and a people problem. And on all three, we are the ones getting in our own way.
Look at the mirror China is holding up, and be honest about what it shows. They are not magic. They made a decision to build and then they built. Last year they added roughly eight times more new power capacity than we did, part of a half-trillion-dollar energy build-out in a single year, and they already generate close to twice the electricity we do. They pulled within about 2.7 percent of our best models on Stanford’s 2026 AI Index, and they did it spending a fraction of what we spend. Admire the execution, because it is genuinely impressive, and then ask the only question that matters: why can’t we?
The answer is not that we lack money, chips, or brains. We have more of all three than anyone. The answer is that we have spent forty years building a country that is very good at saying no, and we are now discovering that you cannot win a build race with a permitting culture.
Start with permits. BNS turns away work not because the demand isn’t there but because we cannot get power to a site fast enough. Interconnection studies take years. A transformer is back-ordered into next year. Customers have the chips sitting in crates and no approved way to plug them in. We did this to ourselves, one well-intentioned review process at a time, and the cost is measured in the one currency this race actually runs on: time.
Then there is our fear. We treat every disruptive technology as a threat first and an opportunity never. We did it with the automobile, with the factory floor, with the internet. Every single time the same crowd warned it would wreck us, and every single time the thing we feared became the floor we now stand on. The pattern is so reliable it is almost funny, and yet here we are again, dressing up fear of the unknown as prudence and calling hesitation a strategy. It is not a strategy. It is a forfeit with better branding.
And it’s personal. We spent a generation telling kids that the only respectable path ran through a four-year degree, and that the trades were what you did if you couldn’t do better. So now, at the exact moment we need to build more physical infrastructure than we have in fifty years, we do not have enough electricians, linemen, and high-voltage technicians to build it. The trade needs something on the order of eighty thousand new electricians a year just to tread water, and hundreds of thousands of skilled-trade jobs already sit open. You cannot code your way out of a transformer shortage. You cannot prompt a substation into existence. Somebody has to actually build the thing, and we spent thirty years quietly insulting the people who can.
So what do we do. Not caution. Abundance. Build everything that makes a clean, reliable electron: nuclear, gas, geothermal, storage, solar, all of it, and stop pretending we have to choose. Bring your own power, so the buildout strengthens the grid instead of straining it. Rip the years out of the permitting process without pretending safety is the reason it takes a decade. And treat the trades like what they are, which is the single hardest constraint between us and winning, and pay and train accordingly.
I will be upfront that I have skin in this game. I build this for a living and I would profit from more of it getting built. Weigh my argument with that in mind. But that is also exactly why I know where the bottleneck actually is, and it is not on the other side of the Pacific. It is in our own habits. It is the review that takes four years, the fear that masquerades as wisdom, and the workforce we talked ourselves out of building.
The competition was never really China. It is our own flinch. We do not get many chances to build something this important. I do not intend to watch us forfeit this one from the sidelines, and if you build this too, I would like to hear how it looks from your seat.
This is the first of a series. I am going to keep writing about what the buildout actually looks like on the ground, where the real bottlenecks are, and how we stop losing a race we are more than capable of winning.
Sources
-
China added roughly 8x more new power capacity than the US in 2025, part of a ~$500B single-year energy build-out. Ember data, via CarbonCredits, 2026.
-
China consumed about twice the electricity of the US in 2025. Bloomberg, citing CCTV, January 2026.
-
The US lead over China narrowed to about 2.7% on the Stanford AI Index 2026, with China competing at a fraction of the investment. Stanford HAI AI Index 2026, via The Next Web, April 2026.
-
US electrician demand runs on the order of 80,000 openings per year, with hundreds of thousands of skilled-trade jobs unfilled. US Bureau of Labor Statistics and industry analyses, 2026.
-
Kimi K3 open-weight release. Moonshot AI, late July 2026 (Bloomberg, CNBC).