China Isn’t Behind. They Might Be Right Behind You
"China is not behind anybody ahead of you. China's right behind us. We're very, very close."
That was Nvidia CEO Jensen Huang. Calm. Certain. Not mincing words.
He said it on camera. And in this game—AI's global race for dominance—his words should chill the room more than any benchmark ever could.
Because here’s the truth: China is catching up. Fast. Possibly faster than most people realise.
You can debate it in theory, or you can just look at the data. The 2024 Stanford AI Index showed US institutions released 40 "notable" models last year, while China produced 15. But that number alone doesn’t tell the story. The real plot twist? Chinese models are closing the performance gap.
Just a year ago, they trailed US models by double-digit margins in benchmarks. Now? We’re talking single digits. Sometimes, no meaningful gap at all.
Take LMSYS Chatbot Arena—a benchmark that ranks models based on human opinion, not just raw technical metrics. Back in January, the U.S. lead looked comfortable. But month by month, that margin has shrunk. From a gulf to a crack. From 150 points to 20.
Sure, OpenAI's GPT-4 and Anthropic’s Claude still hold the crown. But China isn’t building crown-wearers. They’re building contenders. Quietly. Relentlessly.
Case in point: Qwen 3.
Released just days ago, it’s an open-source model out of China that shocked people who actually use these tools day-to-day. Benchmarks? Excellent. Real-world performance? Surprisingly strong. It holds its own against models like Gemini 2.5 Pro on reasoning, maths, and code. It even rivals GPT-4 Omni on some tasks—and it’s free.
And then there's ERNIE X1 Turbo, Baidu’s answer to DeepSeek and OpenAI’s reasoning engines. It punches at the same weight for a fraction of the cost.
So why isn’t everyone talking about this?
Because most people simply don’t know these models exist.
Chinese releases rarely get the marketing polish of OpenAI or Anthropic. Their UIs can be clunky. Their docs? Often untranslated. But if you’re focused on performance, cost, and scalability—and you’re willing to get your hands dirty—these models are becoming real alternatives.
Even ex-Google CEO Eric Schmidt—who said in early 2024 that China was "2-3 years behind"—had to walk that back by year’s end. When DeepSeek dropped, he publicly admitted: the gap is now within a year. Maybe less.
And here’s where it gets serious.
Schmidt didn’t just say this on a podcast. He said it in front of Parliament. Because this isn't just about consumer apps and chatbots. This is about national security.
According to Schmidt and Scale AI CEO Alexandr Wang, if China gets to artificial general intelligence (AGI) first, it won’t just be a tech win. It will be an ideological export. Think Einstein-in-your-pocket, except it's the CCP's Einstein. Embedded in devices. Systems. Warfare. Infrastructure.
Forget tanks. Think swarms of drones, powered by autonomous reasoning, executing attacks with algorithms we haven’t even conceived yet. Think zero-day exploits no Western engineer ever predicted.
Still think the AI race is just another Silicon Valley rivalry?
The U.S. government doesn’t. They’ve started choking off China’s GPU access. First the H100s, then even the downgraded H20s—crippling AI compute at the source. Nvidia took a $5.5B hit just to follow orders.
It might slow China down. For a bit. But scarcity breeds creativity.
What did China do? Built their own chips.
Huawei's Ascend 910C and 910D now power the majority of China’s AI infrastructure. Not yet as efficient as Nvidia's best, but good enough to matter—especially when deployed at scale. Huawei now controls 75% of China's AI chip market. And they’re not stopping.
So, where does that leave us?
The U.S. still leads in usability, polish, and wide adoption. But on raw talent and acceleration? China’s right there. Possibly a year behind. Maybe less.
And this isn’t a 100m sprint. As Jensen said, this is an infinite race. There are no quarters. No final whistles.
Only the next breakthrough. And the next.
So maybe the question isn’t whether China can catch up. Maybe the question is:
What happens when they do?