The truth behind the collective departure of OpenAI executives: Is the father of ChatGPT copying the YC incubator model?

Written by
Caleb Hayes
Updated on:July-11th-2025
Recommendation

What kind of transformation strategy is hidden behind the departure of OpenAI executives?
Core content:
1. OpenAI executives collectively resigned to start their own businesses, but the company chose to invest and support them
2. From the laboratory to the capital game, OpenAI's "overt conspiracy" to build an ecological empire
3. Investors and developer communities expressed concerns about OpenAI's commercialization and closed-source trends

Yang Fangxian
Founder of 53AI/Most Valuable Expert of Tencent Cloud (TVP)

While the world was still marveling at the intelligence of ChatGPT, OpenAI was experiencing a "major earthquake" : Vice President of Post-training Research William Fedus announced his resignation to start his own business , while his predecessor Barret Zoph and others had collectively left half a year ago.

What is even more intriguing is that instead of blocking it, OpenAI chose to invest and support it - is there a hidden ambition behind this that the AI ​​giant wants to transform into an "incubator"?


1. The departure of a talented physicist: from ChatGPT to AI materials science


1.  Fedus’s “cross-border transformation”

This AI master, who was deeply involved in the development of ChatGPT and o1/o3 models, chose an unexpected path:

  • New Track
    Founded an AI materials science company, focusing on the application of "AI for Science" in the field of physics;
  • OpenAI's attitude
    The official statement clearly stated that "they will invest and cooperate with the new company", which can be called "the friendliest departure in history."

Background :

  • Fedus graduated from the Department of Physics at MIT and studied under Turing Award winner Yoshua Bengio for his Ph.D., with a research focus on generative models for reinforcement learning .
  • While at Google Brain, he led the development of the first RL-based text generation model, laying the technical foundation for ChatGPT;
  • After joining OpenAI, he led the team to overcome the post-training difficulties of the o1 series models, achieving a qualitative leap in AI reasoning capabilities.

2.  OpenAI’s “talent black hole”

Fedus's departure is not an isolated case:

  • CTO Mira led the team to escape
    In October 2024, OpenAI CTO Mira Murati founded a new company with dozens of core engineers, and was exposed to have "almost emptied the OpenAI hardware team";
  • The Zoph Era Ends
    Former R&D Vice President Barret Zoph founded an AI chip company after leaving the company, directly competing with Nvidia;
  • Veteran loss
    Co-founder Ilya Sutskever gradually faded out, and CEO Ultraman was accused of "focusing on products rather than research and development."

Industry shock :

  • The flow of talent shows an "OpenAI-style entrepreneurial wave", with startups generally valued at over $1 billion;
  • Silicon Valley venture capitalists bluntly said: "OpenAI is mass-producing AI unicorns, just like the YC incubator."


2. OpenAI’s “overt conspiracy”: from laboratory to capital game


1.  “Free-range” resigned executives to build an ecological empire

OpenAI's "indulgence" in the loss of talent actually has a hidden meaning:

  • Strategic Investment
    By investing in companies founded by former executives, the company can feed back technological achievements to its own ecosystem;
  • Resource Replacement
    Use OpenAI’s brand to endorse startups in exchange for equity gains;
  • Risk Diversification
    Avoid excessive concentration of core technologies and reduce the systemic risks of single model research and development.

Typical cases :

  • Fedus' new company will license OpenAI technology to accelerate the commercialization of AI materials science;
  • Mira's hardware team is developing new AI chips, which may be deeply integrated with OpenAI's cloud computing business in the future.


2.  Ultraman's "Capital Bureau"

CEO Altman's ambitions go far beyond AI research and development:

  • IPO Countdown
    Recently, it was reported that OpenAI was secretly preparing to go public, and its valuation has exceeded $300 billion;
  • Business Diversification
    Launched GPTs store, enterprise-level API, and even entered the field of robots;
  • Global expansion
    A team of 1,000 people is being formed in India with the goal of occupying the Asian AI market within three years.

Focus of the dispute :

  • Investors questioned OpenAI's deviation from its "non-profit origins" and whether excessive commercialization could lead to stagnation of technological innovation;
  • The developer community is worried that the closed-source trend will intensify, and small and medium-sized teams will find it difficult to compete with large companies.



3. “Power Reconstruction” in the AI ​​World: Opportunities and Challenges for Chinese Players


1.  Chinese rivals of OpenAI founders

Fedus's resignation and entrepreneurship coincided with the explosive growth of China's AI industry:

  • The rise of the university group
    Teams from Tsinghua University and the Chinese Academy of Sciences have successively mastered key AI4S technologies, and the number of papers has jumped to the top three in the world;
  • Enterprise distribution power
    Huawei Ascend and Baidu Kunlun chips are accelerating their iterations, benchmarking against NVIDIA H100;
  • The rise of open source ecosystem
    Tongyi Qianwen QwQ-32B topped the LiveBench list, and the number of developers surpassed Meta Llama.

2.  China’s “asymmetric advantage” in AI

Faced with the capital pressure from OpenAI, the Chinese team took a different approach:

  • Policy dividends
    The National Development and Reform Commission explicitly supports the integration of "AI + manufacturing", and local governments have invested hundreds of billions of yuan in subsidies;
  • Scenario Advantages
    The data dividend of 1.4 billion people is in urgent need in the fields of medical care, education, agriculture, etc.
  • Open Source Ecosystem
    Open source models such as Baichuan Intelligence and Zero One Everything cover all modalities and cost only 1/50 of GPT-4.

Expert opinion :

  • Academician Zhang Bei of Tsinghua University: "Chinese AI should focus on breakthroughs in vertical fields and avoid head-on competition with OpenAI on general large models."
  • Partner of venture capital firm: "OpenAI entrepreneurs are good at technological disruption, while Chinese companies are better at commercialization."


4. The future is here: “Dual-track” competition in the AI ​​world


1.  United States: The closed loop of “capital + technology” led by OpenAI

  • R&D model
    Talents from top universities flock to large companies, and technological breakthroughs are quickly transformed into products;
  • Capital Operation
    OpenAI-style incubators give rise to AI unicorn clusters, and valuation bubbles are looming;
  • Risk Warning
    Excessive commercialization may lead to technological loss of control, and OpenAI has been investigated by the European Union for antitrust violations.


2.  China: Breakthrough of “scenario-driven” under the national system

  • Policy driven
    National strategies such as new infrastructure and East-West computing pave the way for AI;
  • Industry Integration
    BYD Super Fast Charging, Xiaomi Audio Inference and other applications set new world records;
  • The open source revolution
    Tongyi Qianwen ecosystem brings together 100,000 developers and is rewriting the global AI discourse.


5. Conclusion: The future of mankind belongs to "controllable creativity"


As OpenAI executives transform into "serial entrepreneurs" and as China's AI army conquers cities and territories, this global AI competition has long gone beyond the scope of technology and evolved into a comprehensive contest of institutional innovation and ecosystem construction .


As Fedus said in his resignation statement: "The future of AI belongs to those who can look up to the stars and keep their feet on the ground."