ChinAI #373: The OpenClaw Hype and Overestimating China’s Diffusion Advantage

Greetings from a world where…

amidst the dearth of late summer TV, one finds himself watching the 1979 British show Tinker Tailor Soldier Spy on Youtube

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: OpenClaw - It rose to fame, it was loved, and now it’s over

Context: Remember when OpenClaw, the open-source AI agent, took China by storm? Tencent and other Chinese tech giants set up offline booths to help people deploy open-source agents. Shenzhen’s electronics market ran low on Mac Minis, which people gravitated to as the most efficient device for running the OpenClaw Agent 24/7. Every day, there was a new breathless report on China’s “raising lobster” frenzy (lobster is a shorthand for OpenClaw).

This reporting and analysis of China’s OpenClaw fever regurgitated a familiar, lazy, and deeply mistaken narrative about China’s diffusion advantage in new technologies. Based on SecurityScorecard data, NBC News reported that OpenClaw usage in China is “now almost double that in the U.S.” Citing this news article, Council on Foreign Relations researchers argued that OpenClaw was a sign of “China’s diffusion advantage.” They wrote, “Cutthroat platform competition, a regulatory imperative to adopt AI, and a growing technologically savvy user base drive deployment at a scale and speed without parallel in the United States.” The point here is not to call out any of these specific reporters or researchers; rather, I aim to undermine this meme of China’s diffusion advantage before it gets entrenched by AI slop beyond the point of no return.

Key Takeaways: This week’s feature translation (a QbitAI article) reminisces about the viral sensation of OpenClaw, which has now become a “thing of the past” [时代的眼泪].

• From November 2025 to March 2026, OpenClaw was one of the fastest-growing open-source projects, as measured by growth in stars. Now, nobody really mentions it.
• Mengyao, the reporter, recalls, “During that period, buying a Mac Mini, installing Clawdbot, and naming one’s Agent became a sort of ‘cyber-home furnishing’ trend within the AI ​​community. Shenzhen’s Huaqiangbei electronics market even faced Mac Mini shortages at one point, with some models selling for up to 600 RMB above list price.” In addition, online platforms offered Lobster deployment services for 499 RMB.

However, regardless of how many Chinese 60+ year-olds were photographed installing OpenClaw, this frenzy was always a misleading indicator of AI diffusion.

• For one, the Lobster required a high threshold for adoption. OpenClaw users had to do a lot on their own: picking the model, connecting the agent to their data and terminals, and integrating agent skills (reusable packages that extend an agent’s capabilities). Shortly after, OpenAI’s Codex and Anthropic’s Claude Code made it much easier for users to work with agents in a “ready-made package.”
• Two other downsides to OpenClaw that limited widespread diffusion: 1) it was token-hungry; 2) its persistent security issues.
• By the way, even if we do take OpenClaw deployment as an indicator of AI diffusion, when I examined at the underlying data from SecurityScorecard (search conducted August 29, 2026), the geographic distribution of OpenClaw instances showed 18.7k deployed in the United States, and 17.0k deployed in China. So, any figure that claims OpenClaw usage in China is double that of the U.S. either misinterpreted the SecurityScorecard data or is very outdated.

OpenClaw is not dead: instead, it has morphed into more accessible versions (one-click deployment via a cloud provider). According to the article, Chinese tech firms like Zhipu AI, Tencent, and ByteDance have launched over 30 derivative products based on OpenClaw.

• In a previous ChinAI issue, we covered one type of OpenClaw-derived product, represented by the MiniMax + Alibaba Cloud alliance.

• JPMorganChase analysts see enthusiasm for OpenClaw and AI agents like AutoClaw (Zhipu) and MaxClaw (MiniMax) as providing a short-term boost for Chinese technology stocks; still, in their revenue forecasts for major Chinese cloud service providers, the revenue growth rate is projected to decline from 2025 to 2026, and decline again from 2026 to 2027.
• If diffusion of AI to Chinese businesses will largely travel through the cloud (as this OpenClaw case points toward), then it’s very likely China will suffer from strong diffusion deficit compared to the United States. As I pointed out in this article titled “China’s AI Implementation Gap”, China’s cloud computing adoption significantly lags that of the United States.

FULL TRANSLATION: OpenClaw: It rose to fame, it was loved, and now it’s over

Ask the Audience: tricky research ethics question

I know many readers have a lot of China studies expertise and training, so I wanted to ask the audience about a tricky research ethics issue I’ve encountered. I’m working on an early-stage project that investigates Chinese assessments of power, and I came across a report published over 10 years ago on the subject. The report contained some fascinating quotes from Chinese scholars, but when I double-checked the original sources from the China National Knowledge Infrastructure database, I could not find the quotes.

I’ve reached out to the report’s author, and, in our discussion, we landed on two charitable interpretations for why this discrepancy exists: 1) for one quote, the author relied on the Open Source Center translation (a U.S. government service that used to translate foreign-language publications), so maybe the OSC translation was incorrect; 2) another possibility is that the original Chinese sources have been edited/censored in the time between when the author quoted them in this old report and when I checked on them recently.

My question for those who work with Chinese translations and conduct research in this space: have you encountered these two issues before? Both of the original Chinese sources are articles published in Contemporary International Relations [现代国际关系], a relatively high-profile publication, so maybe certain articles get edited/censored after their publication.

What next steps would you pursue, if you were in my situation? The report’s author doesn’t seem interested in revising or correcting an old report, and to my knowledge, think tanks might not have established processes for issuing corrections like academic institutions. I’m not interested in a “gotcha”-type witch hunt, as this is an older report. However, I’ve seen these quotes get cited by other high-profile researchers, and I could have easily incorporated them into my own research, if I hadn’t gone back and double-checked. It also would be genuinely helpful for my own research, if it would be possible to confirm that those quotes did exist and were later censored and cut out.

Thank you for reading and engaging.

These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

It’s worth noting that each of these three factors do not support a diffusion advantage for China. Competition occurs everywhere, and leads to more organic and sustainable diffusion in market-based economies; AI companies probably encounter more regulatory obstacles in China than in the United States; and if diffusion is about the entire economy, China’s has much less technologically savvy user base than the United States.