More than 1.5 billion salaries can't be kept! AI Chinese genius Pang Ruoming changed jobs twice in 7 months. Why did she abandon Meta and switch to OpenAI?
The battle for artificial intelligence (AI) talents is intensifying, and OpenAI has successfully cornered Ruoming Pang, a well-known AI researcher, from Meta. According to the U.S. media The Information on Wednesday (25th), Pang Ruoming left Meta last week and officially joined OpenAI, which is his second job change in about seven months. The OpenAI spokesman has confirmed the news. It is reported that the company has been cornering Pang Ruoming for several months. Although he once expressed his work in Meta, he finally decided to leave. Bloomberg reported that when Pang Ruoming joined Meta in 2025, the total salary plan for several years exceeded 200 million US dollars (about 1.56 billion Hong Kong dollars, the same below).
Abandon Meta's salary of more than 1.5 billion
Pang Ruoming, a Chinese scientist who once dominated the underlying technology of Apple's artificial intelligence and was later cornered by Meta with a sky-high salary of more than 200 million US dollars, finally chose to switch to his competitor OpenAI after only seven months of joining Meta. According to the source, OpenAI spent several months digging fiercely and finally successfully absorbed the expert who is known as the "top master" in the field of AI infrastructure.
Pang Ruoming has a deep background in both academic and industrial circles. He received a bachelor's degree from Shanghai Jiaotong University, and then went to the United States for further study and received a doctorate from Princeton University. After completing his doctoral studies, he joined Google for 15 years. During his tenure at Google, he was the core developer of the deep learning framework Lingvo and made key contributions to the design and development of the framework. After joining Apple in 2021, he built the technical foundation of "Apple Intelligence" from scratch and led the development of the AXLearn framework for functions such as email digests and Genmoji.
In July last year, Zuckerberg, CEO of Meta, personally came out to recruit him with a compensation plan with a total value of more than 200 million US dollars, including basic remuneration, signing bonuses and huge options, to be responsible for the infrastructure work of the super intelligent laboratory. Although Pang Ruoming once said that he had a good job at Meta, he finally refused OpenAI's sincere invitation and gave up his huge wealth to choose to leave.
Why choose OpenAI?
What do people who can't keep 200 million dollars want? The answer is in the layout of OpenAI. Some industry insiders said bluntly that what really impressed him was 'comulation power + Sora-level world modeling infrastructure'. Without these two, it is almost impossible to create a top system in 2026.
Pang Ruoming specializes in AI infrastructure. He is a rare talent who understands both technology and is good at construction. He can achieve the performance of 15,000 GPUs with 10,000 GPUs. In Meta, he may be able to optimize the existing infrastructure; but if you want to explore the next generation of AI, such as physical intelligence and world simulation, OpenAI is the most suitable platform.
Sora's success is not only the victory of the algorithm, but also the overwhelming advantage of the underlying architecture. With its leading position in the field of AGI and a strong research and development platform, OpenAI has shown a great magnetic attraction effect on top talents. When engineers find that the system they have carefully built is far inferior to the top architecture of the other party, leaving to pursue technical limits is not a betrayal, but an inevitable choice. For Pang Ruoming and other talents, money is no longer the first, and the technical upper limit brought by the platform is the key.
Meta's brain drain
Pang Ruoming's lightning departure is just the latest example of the brain drain of Meta's core AI team. Last week, Mat Velloso, the product manager of Meta Super Intelligent Laboratory Developer Platform, left shortly after taking office; Russ Salakhutdinov, vice president of generative AI research, also announced his departure after a two-year term; earlier, he served as Meta's chief AI for a long time. Scholar Yann LeCun not only "retreated to the front line", but also officially left Meta and founded his own company (AMI Labs) from the end of 2025 to the beginning of 2026.
According to incomplete statistics, many of the more than 50 top talents that Meta dug from OpenAI, DeepMind and other heavy money last year have left one after another. Engineer Avi Verma repented after completing the onboarding procedures and returned to OpenAI. This reflects the awkward situation of Meta: since the launch of Llama 4, although it has been greatly expanded, it has never been able to launch subversive products comparable to GPT-4o and Sora. For top researchers, the most worrying thing is not the hard work, but the hope of the invisible results.
The battle for AI talents in Silicon Valley is almost crazy. The industry pointed out that Silicon Valley is still the craziest casino on the road to AGI, while OpenAI temporarily holds the largest card table. Whether Meta can turn over with the Llama 4 model remains to be verified by time. After all, in Silicon Valley, loyalty only exists before the next killer product appears.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
