Will China Ever Catch Up in the AI Race? Top Chinese Experts Say 'Don’t Hold Your Breath'

中国真能赶上美国的AI发展速度吗？顶尖专家回应：别抱太大希望

www.bloomberg.com

So much for the 'China is dominating AI' hype. Justin Lin from Alibaba’s Qwen team just put the odds of a Chinese AI breakthrough at under 20% in the next 5 years. That’s not just cautious — it’s a reality check from someone actually building the models.

之前炒作‘中国AI领先全球’的声音可以歇歇了。阿里巴巴通义千问团队的林俊却表示，未来五年内中国在AI上取得重大突破的概率不到20%。这不只是保守估计，更是来自一线模型研发者的现实警告。

And it’s not just him. Tencent and Zhipu AI execs echoed the same sentiment — especially after Zhipu’s IPO. The message is clear: homegrown innovation is still playing catch-up, not leading the league. If even the builders are this sober, maybe we should all stop cheering the ‘AI miracle’ narrative.

而且不只他一个人。腾讯和智谱AI的高管也表达了相似观点——尤其是在智谱上市之后。信息很明确：本土创新仍在追赶，而非领跑。如果连建造者都如此清醒，也许我们都该停止鼓吹‘AI奇迹’的叙事了。

Venture Capitalist from Beijing (北京风险投资人)

People keep treating AI like it’s just code and data, but the real bottleneck is access to cutting-edge chips. The US isn’t just ahead — it’s controlling the supply chain. You can have all the genius engineers in the world, but if you can’t run the models, it’s academic.

很多人把AI当成只要代码和数据就能搞定的事，但真正的瓶颈是先进芯片的获取。美国不只是领先，它还掌控着整条供应链。你就算拥有全世界最优秀的工程师，如果跑不动模型，一切也只是纸上谈兵。

Grad Student in Silicon Valley (硅谷研究生)

This isn’t surprising. I see the compute gap every day in research labs. Chinese teams often have to optimize models just to run on weaker hardware. Meanwhile, US labs train trillion-parameter monsters without blinking. The asymmetry is breathtaking.

这并不意外。我在研究实验室每天都能看到计算能力的巨大差距。中国团队常常被迫优化模型以适应较弱的硬件，而美国实验室却能毫无压力地训练上万亿参数的巨兽模型。这种不对称性令人震惊。

Hardware Hacker from Guangzhou (广州硬件极客)

But can’t we just build our own chips? We’re not exactly amateurs at semiconductor design.

但我们不能自己造芯片吗？我们在半导体设计上又不是新手。

Policy Analyst from Hong Kong (香港政策分析师)

Let’s be real: chipmaking isn’t just engineering. It’s geopolitics, trillions in capital, and decades of ecosystem buildup. You can’t ‘hack’ ASML’s lithography machines in a garage.

现实点吧：芯片制造不只是工程问题。它涉及地缘政治、数万亿美元资本投入，以及几十年来的生态积累。你不可能在车库里黑掉ASML的光刻机。

Skeptical Tech Blogger (怀疑派科技博主)

Honestly? The ‘China vs US AI race’ narrative is overblown. AI isn’t a sprint — it’s a relay with a hundred legs. Different countries will lead in different areas. China’s strength in deployment and scale shouldn’t be ignored.

说实话？‘中美AI竞赛’的说法被夸大了。AI不是短跑，而是由上百段组成的接力赛。不同国家会在不同领域领先。中国在落地应用和规模化上的优势不应被忽视。

Optimistic Startup Founder (乐观的初创公司创始人)

Breakthroughs don’t happen on schedules. Remember how everyone counted China out in 5G? Now Huawei’s leading globally. Desperation breeds innovation. Sanctions might just be the push we need.

突破不会按计划发生。还记得当初所有人都不看好中国在5G领域的表现吗？如今华为已在全球领先。绝境催生创新。制裁或许正是我们需要的推动力。

Cynical Ex-Engineer (愤世嫉俗的前工程师)

Here we go again. Another PR-friendly ‘reality check’ right before a funding round. Of course they say it’s hard — they want more subsidies. Show me the open benchmarks, not the press releases.

又来了。每次融资前都来一次公关友好的‘现实警告’。他们当然说很难——是为了争取更多补贴。给我看公开的基准测试，而不是新闻稿。

Will China Ever Catch Up in the AI Race? Top Chinese Experts Say 'Don’t Hold Your Breath'

中国真能赶上美国的AI发展速度吗？顶尖专家回应：别抱太大希望

中国的机器人已经上岗了，而美国还在做演示？我们正输掉未来吗？

AI泡沫？不，我们才刚刚踏入危险区域——桥水基金顶级投资人如是说