The US claims it’s putting China “in its place,” but its actions are anything but! What does H200’s return to the Chinese market signal?

“Restrict China’s AI development!”
“Approve Chinese companies to purchase H200!”
If you put these two statements together, people who don’t know might think the US is schizophrenic.
But reality is just that surreal. Recently, the US approved some Chinese companies to purchase Nvidia H200 chips, and the global tech community immediately went into a high-energy frenzy. Because this means the AI chip war is entering a new phase of “fighting while selling.”
H200 is not an ordinary GPU—it’s top-tier equipment in the AI world. Training large models, running complex reasoning, and generating AI videos all rely on this kind of high-end computing power. Put simply, whoever controls more H200 gets one step closer to becoming the AI overlord.
So the moment the news broke, Chinese companies reacted faster than food delivery riders grabbing orders. That’s because everyone understands the rules of the market too well: being able to buy today doesn’t mean you’ll be able to buy tomorrow.
Of course, Nvidia is the happiest. Over the past two years, Jensen Huang has almost been “lifted onto the pedestal” by the capital markets. The more popular AI becomes, the more GPUs are in short supply; the more GPUs are scarce, the more expensive Nvidia becomes. Now that the US has loosened its stance just a bit, it’s like opening another vault door for Nvidia.
And the US has its own calculations too. If restrictions are too harsh, Chinese companies will go into a frenzy of self-developed innovation; if it’s fully opened, they worry the technological lead will shrink. So they’ve adopted a “controlled release” strategy—like a faucet: turn it on a little, but not all the way.
This move may look smart, but the market can only produce one result: companies worldwide will start hoarding chips even more aggressively.
Because uncertainty itself is the biggest stimulant.
Many companies are no longer buying “just when needed”—they’re buying “as much as possible as quickly as they can.” In the past, warehouses stored raw materials; now they store GPUs. In the AI era, the hardest asset isn’t land anymore—it’s computing power.
Even people have joked: in the future, landlords might not collect rent in cash, but instead collect H200.
But the truly dangerous part is the escalation of industry-wide competition. As more high-end chips flow into the market, competition among large models will only intensify. AI application companies will enter a “burn computing power to win users” frenzy—just like the internet days when companies burned money to capture market share.
The difference is that back then they burned subsidies; now they’re burning electricity bills.
This H200 incident, at its core, isn’t really a chip story—it’s a snapshot of how the global AI landscape is shifting. The US can’t completely give up the China market, and China can’t stop chasing the lead.
The end result is this: compete while collaborating; restrict while still profiting.
And the busiest people might not be engineers anymore, but global logistics companies. After all, everyone is waiting for the same question: Has the chip arrived? #Gate广场五月交易分享