Forget DeepSeek. That’s yesterday’s AI disruption, and it’s out of style already. Everybody’s wearing Qwen now.

Wait, what’s “Qwen”? Qwen’s Alibaba’s disruptor candidate, and news that its open-source model, QwQ, can go toe-to-toe with DeepSeek using far less in the way of data looked like yet another kick in the shins to America’s vaunted hyper-scalers, who still generally insist the only real and true path to the AI dystopia none of us were allowed to vote on goes through unbridled spending in pursuit of ever larger compute capacity.

As VentureBeat explained, “QwQ achieves comparable performance [to DeepSeek-R1] with a much smaller footprint.” That sounds bad for OpenAI, and for big US tech more generally.

If Alibaba’s now the standard-bearer for AI disruption, it’s a testament to the idea that the company’s back in a big way after suffering through a nightmarish several years tied to Jack Ma’s ill-advised decision to rankle Xi Jinping in late 2020.

The stock rose another 8% in Hong Kong Thursday. It’s up 70% or so this year, adding some $150 billion in market cap in the process.

We’re now entering what feels like a Chinese AI hype cycle which Beijing’s for now willing to facilitate as a means of restoring market confidence.

At the NPC this week, the Party tipped a keen interest in supporting AI R&D, and if you had to say what Xi’s top priority is right now, self-sufficiency in high-tech areas of economic development would be a good guess.

Note that the Hang Seng Tech gauge has outperformed the Nasdaq 100 in seven of nine weeks so far this year.

The Qwen news was insult to injury for US tech on Thursday, when Marvell plunged a God-awful ~19% after failing to impress markets with its current-quarter top-line guide. The shares are now destined to notch a seventh straight weekly decline. The stock’s down 40% from the highs.

As Deutsche Bank’s Jim Reid pointed out in a recent special report, “today’s AI capex boom is not being financed by debt” but rather primarily out of US mega-cap earnings, which ostensibly “reduces systemic risk.”

And yet, the recent run-up in US household wealth’s attributable in no small part to the rally in AI hyper-scaler names. That, in turn, means there are significant domestic macro risks around a pullback for the market leadership. And as of noon Thursday, big US tech was flirting with a correction.

The figures illustrate the point. “The market has never been so concentrated in terms of exposure to the largest stocks that are heavily investing in AI capex,” Reid wrote, adding that “if we do see a temporary AI winter, it could dramatically impact wealth in the US and could disrupt the economy even if a destructive debt unwind is highly unlikely.”

Importantly — and I’ve mentioned this repeatedly since the DeepSeek shock — the risk isn’t so much that interest in AI will wane as it is investors will question assumptions about the correlation between spending and innovation. Put differently: The risk isn’t an “AI winter,” but rather a bursting of the US AI capex bubble.

As DeepSeek founder Liang Wenfeng put it last year, “More investment does not necessarily lead to more innovation.”