Jordan Schneider on China's $9K EVs, astroturfed data center opposition, DeepSeek's Liang Wenfeng, and LARP rockets

Jordan Schneider on China's EVs, data center opposition, and LARP rockets

Jordan Schneider, founder of ChinaTalk, covers four threads: Chinese EV pricing, the real roots of data center opposition, DeepSeek's Liang Wenfeng, and the corruption scandal hollowing out China's Rocket Force.

---

$9,000 EVs

Chinese EVs are now selling for around $9,000 — a price Schneider describes as feeling indistinguishable from a $35,000 car. The economics are straightforward for markets that haven't closed their doors: countries like Malaysia and Indonesia get access to a vehicle at roughly one-third the cost of Western alternatives, often with better reliability because EVs have fewer moving parts to fail. Europe has responded with tariffs; the US has effectively banned them. For everyone else, Schneider says, it's a no-brainer consumer win.

---

Astroturfed data center opposition

The framing that Chinese influence operations are driving data center protests in the US doesn't hold up. The actual evidence — an OpenAI report identifying influencers using VPNs and ChatGPT to generate anti-data-center images — amounted to roughly 200 posts that averaged three likes each. China has a long, well-documented track record of being bad at US social media influence campaigns. Schneider taught himself Python in 2019 to analyze a Twitter bot network takedown and found the posting schedule mapped almost exactly to Beijing working hours, including a nap break from 1 to 3PM.

None of that means the opposition is manufactured. There are completely legitimate reasons ordinary Americans would show up to a city council meeting against a data center — concerns about water use, power draw, noise, diesel backup generators, and the general sense that these facilities represent companies that want to automate their jobs. The foreign influence story is a distraction from a domestically generated grievance.

---

“They're really cheap. That's just the crazy thing... it's just wild to have a $9,000 car, which feels exactly like a $35,000 one... The evidence people are pointing towards is this one OpenAI report where they like caught these influencers using a VPN and using ChatGPT to generate images saying data centers suck. Those things got like three likes each... Liang Wenfeng: he's told his investors, I'm not here to make a profit. I'm here to make AGI.”

Liang Wenfeng and DeepSeek

Liang Wenfeng is the closest analog to Demis Hassabis at DeepMind before the Google acquisition — a founder who is genuinely not optimizing for a business model. DeepSeek is funded by his quant hedge fund, he has told investors he is not there to make a profit, and he has explicitly kept headcount small to avoid the perverse incentives that come with a large commercial operation. He is not beholden to a compute provider or a platform like Alibaba or Tencent.

The leaked transcript that caused a stir was never meant to go public. Liang was speaking privately, and when it leaked he pulled his IPO. What was striking in it, according to Schneider, wasn't the 2050 speculation — it was the clarity about wanting to stay independent and prioritize fundamental research over a product roadmap. That kind of candor is rare enough in China that even Liang, a billionaire with substantial political cover, only expressed it in what he thought was a closed room.

On chips, DeepSeek is sourcing as much Nvidia as it can get. The structural problem for all Chinese AI firms is a 12-to-15x capacity gap between the Western semiconductor ecosystem (Nvidia, TSMC) and the Chinese one — a ratio Schneider says is traceable through CapEx spend, equipment acquisition, and yield rates, and is unlikely to close within the next five years. Chinese AI companies going public are generating $500 million to $1 billion in revenue, which looks fine until you set it against the tens of billions the US frontier labs are mobilizing. The reason is simple: they are not on the frontier, so they cannot charge a premium, and they are compute-constrained.

Moonshot AI (Kimi) is attempting a workaround — licensing its models to Western cloud providers like Microsoft and Azure for a revenue share — but whether the Trump administration will allow that to continue is an open question.

---

LARP rockets

China's Rocket Force — the branch responsible for ground-launched missiles — has been running a multi-year corruption purge after a Bloomberg report several years ago revealed that a significant share of the missiles were filled with water rather than fuel. Two generations of Chinese military leadership have now been purged. The joint chiefs equivalent has gone from nine members down to roughly one and a half.

What makes the situation unusual is that the government has now opened a public tip line asking citizens to report procurement violations. For a system built on internal discipline and opacity, going to the public for whistleblowers suggests the internal audit mechanisms have failed to root out the problem after years of trying. As Schneider notes, checking whether a missile contains fuel or water is not technically difficult — the fact that outside tips are now being solicited says something about how thoroughly the corruption was embedded.