Chen Tianqiao's Overseas AI Island and Dai Jifeng's Open-Source Awakening: The MiroMind Dispute Explained

Forum topic · ✨步子哥 · 2026-04-29

Summary

This Chinese forum post examines the fallout between billionaire Chen Tianqiao, founder of Shanda Group, and AI researcher Dai Jifeng over their startup MiroMind. Chen, once China's richest man after Shanda's 2004 Nasdaq listing, later pivoted to brain science and in March 2025 partnered with Tsinghua associate professor Dai Jifeng to found MiroMind, an open-source large-model company targeting DeepSeek and Neuralink-adjacent goals. Their collaboration collapsed when Dai resigned in January, negotiated a 15% equity stake with rights to MiroMind's IP, then saw his backers demand the stake drop to 5% with unconditional global IP rights. MiroMind rejected the terms; Dai went public via The Washington Post, and MiroMind issued an internal notice accusing him of breaching business integrity. The dispute is rooted in China's 2025 decision to block Meta's $2 billion acquisition of Manus, after which Chen told Bloomberg that MiroMind had built an "information wall" separating its China and US AI operations. Dai then founded Naiveai, reportedly valued around $800 million. The post argues that geopolitical "choosing sides" is now a survival reality for AI startups, and that real technology delivery matters more than media spectacle.

Chen Tianqiao: From Nasdaq Billionaire to Brain-Science Investor

Chen Tianqiao, once China's richest man, led Shanda Group to a 2004 Nasdaq listing that raised $152 million, making him a billionaire at thirty. After losing ground to Tencent and Alibaba, Shanda privatized in 2012 and became an investment group. Suffering from panic disorder, Chen flew from China to California only once and has lived overseas for sixteen years — first in Singapore, then California.

He founded the Tianqiao Brain Research Institute and TCCI, pursuing non-invasive brain-computer interfaces. As he put it: on the brain-computer revolution, "Musk goes left, I go right" — implanted chips versus non-invasive approaches.

ChatGPT Sparks a New Venture: MiroMind

The generative AI wave convinced Chen that AI could accelerate brain science. In March 2025, he partnered with Dai Jifeng, an associate professor at Tsinghua University's Department of Electronic Engineering, to found MiroMind, focused on open-source large models — benchmarking against DeepSeek while supporting brain-science goals.

Dai is a leading computer-vision scholar: co-creator of Deformable Convolutional Networks and the multimodal foundation model InternVL, with over 50,000 citations. He previously served as principal researcher at Microsoft Research Asia and executive research director at SenseTime before returning to academia in 2022. MiroMind's flagship search-agent model, MiroThinker 1.5, has been released.

The Breakdown: Equity and IP Dispute

Dai joined in March 2025 and requested to work in the US; MiroMind sponsored his O-1 visa, approved in November.
On January 15, Dai proposed resignation, offering 15% equity in exchange for legally using MiroMind's IP and bringing core domestic team members.
MiroMind agreed, even helping connect him with investors including IDG and Sequoia.
On March 2, Dai's backing investors demanded the stake drop to 5%, plus an "unconditional, global, perpetual, irrevocable, sublicensable" free IP grant. MiroMind refused.
On April 22, Dai told The Washington Post that after the Manus incident MiroMind pressured him to develop overseas, which he declined.
On April 24, MiroMind issued an internal notice accusing Dai of breaching business integrity, saying it holds leads on trade-secret infringement and reserves rights to criminal complaints and civil litigation.

The Manus Trigger and China's Regulatory Response

The root cause: Meta's $2 billion acquisition of Manus. Although registered in Singapore, Manus's founding team originally developed its technology in China. Chen invested $100 million in MiroMind, part of Shanda's $2 billion "Discoverable AI" investment total.

On April 27, China's National Development and Reform Commission (NDRC) prohibited the acquisition, ordering the deal revoked. CCTV commented that the ban targets "shower-style" offshore restructuring and security risks within openness, not openness itself.

On April 28, Chen told Bloomberg that MiroMind had implemented agreements prohibiting cross-border sharing of information or code, minimizing flows of personnel, data, and assets — a reform following regulatory inquiries:

> "I used to believe we could bring together Chinese and international AI talent for humanity's future. But after the Manus incident, we had to implement full risk control. This is like 'cutting off our own retreat,' yet it is a necessary compromise. Under the geopolitical situation, companies have no choice but to pick a side."

The "Information Wall" and the Aftermath

In March, Chinese officials warned Chen's team against unilateral technology transfer. Chen proposed regional isolation — each region's business managed locally — resolving the issue. MiroMind plans its first external fundraising in the second half of the year, with revenue from asset management and energy infrastructure transactions.

Dai subsequently founded Naiveai, raising roughly $300 million at an ~$800 million valuation, with a core team including former MiroMind members.

Commentary: Substance Over Spectacle

The author (who discloses no shareholder ties or interests with Chen) argues:

One-sided online pile-ons and deliberate PR are unproductive noise.
What matters is whether core technology reaches real mass production and whether teams can close the commercialization loop.
Reports noted MiroMind claimed Dai had Chinese AI researchers remotely serving a Singapore company using US AI chips and compute, facing cloud-control legal risks; Chen's caution was reasonable, and Dai's position was not wrong either. Both sides have a point.

> Geopolitical "choosing sides" is not a simple political statement but a survival calculation. Like Silk Road caravans forced to pick a route at a checkpoint, AI startups must consider markets and exit channels from day one — selling to foreign tech giants now carries high policy risk.

Closing

Chen's brain-computer dream and Dai's open-source path may have diverged, but both aim at advancing human intelligence. The author hopes both sides will turn conflict into progress and let technology genuinely benefit society.

---

References 1. Chen Tianqiao's statement to Bloomberg on MiroMind risk controls and the Manus incident, April 28, 2026. 2. MiroMind internal notice regarding Dai Jifeng's alleged false statements, April 24, 2026. 3. NDRC decision prohibiting Meta's acquisition of Manus, April 27, 2026. 4. Yicai and Caixin reports on MiroMind/Naiveai equity and departure details, 2026. 5. Bloomberg in-depth interview with Chen Tianqiao on MiroMind investment and the AI information wall, 2026.