Author: Xinxin
OpenAI's Vibe Coding dream has been shattered.
On July 11 local time, it was reported that Google DeepMind successfully 'recruited' the core team of the AI startup Windsurf. Not long ago, OpenAI was negotiating to acquire Windsurf for $3 billion, and Geek Park was discussing it on a podcast. Unexpectedly, the cooperation between the two parties did not materialize, but instead allowed Google to replenish its AI talent.
According to reports, Google will pay $2.4 billion (approximately 17 billion yuan) in licensing fees and compensation to bring Windsurf co-founder Douglas Chen and some senior researchers to Google, helping the latter with its AI programming projects. Meanwhile, Windsurf will continue to operate independently and can still license its technology to other companies.
Familiar formula, familiar taste.
Just a month ago, Meta did something similar—Meta invested heavily to acquire nearly half of Scale AI's shares and subsequently brought its young CEO on board as its chief AI officer.
Whether it's Meta, Google, Apple, or Musk's xAI, all are now competing for talent, either incorporating entire star startup teams or directly 'poaching' from OpenAI and Anthropic.
Major giants are luring talent with salary packages worth tens of millions or hundreds of millions of dollars, quickly 'blasting' to poach rival teams, with CEOs personally making calls, organizing meetings, or investing in acquisitions just to secure a few founders and key technical personnel, while the poached rivals are forced to use higher retention bonuses to 'stop the bleeding' and retain staff.
It can be said that the 'AI talent war' in Silicon Valley has reached a frenzy, with 99% of the money ultimately flowing to 1% of top AI talent.
01 Meta is crazily throwing money, draining competitors
Among all the giants, Meta and Zuckerberg's style of poaching may be the most high-profile and aggressive.
In June this year, Meta reorganized its AI team, officially announcing the establishment of the 'Super Artificial Intelligence Laboratory' and invested $14.3 billion to acquire 49% of the shares of data labeling startup Scale AI, directly appointing this young company's CEO, Alexandr Wang, as Meta's chief AI officer, making it a case of 'buying a company and getting an executive.'
Alexandr Wang with Zuckerberg | Image source: Internet
In addition to 'buying talent' through investment firms, Meta is also very generous in its offers to individual talents, especially targeting top researchers from OpenAI and Google, as well as Apple and Anthropic.
Originally, those individuals commanded annual salaries of several million to tens of millions of dollars, along with stock options, already considered top in the industry. To lure away core members from OpenAI, Meta did not hesitate to offer packages at the '4-year $300 million' level, with a substantial amount of stock vesting in the first year, cashing out $100 million. Although Meta claims these extreme offers are limited to 'a few leadership positions,' they are still unprecedented in the tech circle.
Correspondingly, OpenAI, which possesses top AI models, has become the biggest target for poaching; it can be said that there has been severe talent loss, almost turning into an 'AI talent supermarket,' with various giants shopping around, and Meta has at least poached 7 top researchers and model developers from OpenAI.
An executive from OpenAI described being poached by Meta as 'someone breaking into our house to steal things.' Sam Altman certainly feels that the situation is not good, but claims that none of OpenAI's 'best' employees have been poached.
'Meta has started offering huge contracts to many people in our team,' Sam Altman said in June this year on his brother's podcast, 'For example, a signing bonus of $100 million a year, with even higher salaries.' 'But so far, among our best employees, no one has decided to accept his terms.' Meta is said to have attempted to poach one of OpenAI's chief researchers and an AI architect from Google, but both attempts failed.
Sam Altman also sarcastically stated that Meta is obsessed with providing high salaries for employees instead of realizing the mission of AGI, which may not create 'a good culture.'
Sam Altman discusses the Meta poaching incident on a podcast | Image source: Internet
Even so, in addition to using so-called 'culture' and 'vision' to retain staff, OpenAI still has to pay a price in the talent poaching battle, needing to adjust salaries and offer retention bonuses of $1 million to $2 million to some employees, along with more stock options as loyalty rewards, persuading key researchers to stay after receiving external offers. It is said that the board crises and organizational turmoil over the past two years have also affected OpenAI employees' sense of belonging, making some competitors and headhunters feel that 'it is easier to poach from OpenAI.'
However, OpenAI is not just passively taking hits; it has been poaching talent or counter-poaching as well, not only bringing back a researcher from Meta but also stealing a senior VP and several core engineers from Musk's xAI and Tesla, some of whom participated in building Musk's supercomputer Colossus. Musk and Sam Altman have long had a falling out due to differences in direction and are even suing each other now.
In such a salary environment, even the traditionally aloof Apple had to change its style. Apple originally did not encourage researchers to publish papers due to its culture of secrecy, making it difficult to attract top AI scholars. In contrast, companies like Google, Meta, and Microsoft have long allowed researchers to publish papers and open source some tools, which can increase their influence. By 2025, Apple seems to have begun to relax some restrictions and is heavily investing in internal large model projects.
Even so, the supervisor responsible for foundational model research at Apple was still poached by Meta with a package exceeding $100 million; this compensation rumor even surpassed all Apple executives except CEO Tim Cook, and Apple did not attempt to counter-offer or match Meta's offer.
To compete for AI talent, some companies are also adjusting measures, shortening vesting periods; for example, Google has reduced the vesting period for some AI recruitment from 4 years to 3 years to enhance salary attractiveness. Signing bonuses in the tens of millions of dollars are also not uncommon.
It is also worth noting that the offers made by the giants are not purely high salaries but also include stock options and one-time signing bonuses. Some offers are said to carry 'explosion deadlines'—to be signed within 24 hours, otherwise they become void.
Currently, in the entire Silicon Valley AI circle, some resumes resemble a transfer market: having worked at Google, moving to OpenAI for a promotion, then being poached by Meta, and perhaps later starting a new company based on previous experiences to secure hundreds of millions in financing. Of course, some people choose to bounce back and forth; for example, an engineer poached by xAI returned to OpenAI in less than a year. Others choose to reject Meta's high salary package simply because they do not want to 'compete.'
As Meta has raised signing fees to 'superstar' levels, an image comparing a Chinese AI researcher with football superstar Cristiano Ronaldo has even become a popular meme in the tech circle.
Netizens compare Chinese AI researchers with Cristiano Ronaldo | Image source: X
02 Chinese faces become 'hot commodities'
In this fierce talent poaching battle among giants, if you pay attention to their names or surnames, you'll notice quite a few people of Chinese descent.
For example, Jiahui Yu, who is pictured next to Cristiano Ronaldo in the previous image, is of Chinese descent, a graduate of the Youth Class of the University of Science and Technology of China, previously worked at Google DeepMind leading the Gemini multimodal project, then joined OpenAI to participate in the development of models like GPT-4o, GPT-4.1, o3, o4-mini, before being heavily poached by Meta.
The Chinese talent poached by Meta is also a Chinese, named Ruoming Pang, for whom Meta reportedly offered a total package exceeding $200 million to lure him. Ruoming Pang worked at Apple for four years, leading the foundational model team for Apple's AI/machine learning, primarily developing the foundational models supporting Apple Intelligence.
Before working at Apple, Ruoming Pang worked at Google for 15 years, during which he participated in speech recognition research and product development, co-developed the Babelfish/Lingvo deep learning framework and Tacotron 2 speech synthesis system, and was a co-founder and technical leader of Google's global authorization system Zanzibar.
Ruoming Pang | Image source: His X account
In addition to Ruoming Pang, Meta's list of poached Chinese talents includes several former Chinese researchers from OpenAI and Google, who previously participated in the development of cutting-edge large model versions like GPT-4, Gemini, and o-series.
For example, Huiwen Chang is a graduate of Tsinghua University's Yao Class, earned her Ph.D. from Princeton University, and worked for over four years as a research scientist at Google, inventing the MaskGIT and Muse architectures. In 2023, she joined OpenAI and contributed to the development of the GPT-4o image generation system in multimodal AI models.
Huiwen Chang | Image source: Linkedin
For example, Hongyu Ren graduated from Peking University for his undergraduate studies and earned his Ph.D. from Stanford University, previously interning at Microsoft, NVIDIA, Google, and Apple. After joining OpenAI, he was responsible for the post-training team, focusing on optimizing language model training, and was one of the developers of models like GPT-4o mini and o1-mini.
Hongyu Ren | Image source: His personal website
There is also Ji Lin, who graduated from Tsinghua University for his bachelor's degree and from MIT for his Ph.D., joining OpenAI in 2023 as a technical team member, contributing to the development of GPT-4o, GPT-4.1, GPT-4.5, the image generation system (4o-imagegen), and the Operator reasoning stack.
Ji Lin, poached by Meta from OpenAI | Image source: His personal website
In July 2025, Google announced it would hire the CEO and co-founder of AI coding startup Windsurf, along with some R&D employees, integrating them into the Google DeepMind team, thwarting OpenAI's acquisition plan for Windsurf.
Among them, Douglas Chen, a co-founder of Windsurf, who was recruited by Google, is also of Chinese descent, graduated from MIT, and previously worked as a machine learning engineer at Meta and Facebook.
Douglas Chen, co-founder of Windsurf | Image source: Linkedin
Apple also relies on some Chinese talents in AI. After Ruoming Pang left, Apple quickly promoted another Chinese engineer, Zhi Feng Chen, to continue overseeing the research and deployment of large language models behind Apple Intelligence.
The high density of Chinese faces is not a coincidence. According to some think tanks' analysis of authors of top AI conference papers worldwide, over 30% of top AI researchers in the U.S. have a Chinese background, a proportion even slightly higher than those born locally in the U.S.
Musk's preference for Chinese engineers is also very evident; his team photos often feature many Chinese faces, and even at the founding of xAI, out of the 12 founding researchers, 5 were of Chinese descent—Tony Wu, Jimmy Ba, Greg Yang, Zihang Dai, Guodong Zhang—many of whom had interned or worked at Google or DeepMind, or OpenAI. Sometimes people jokingly say, 'Half of xAI is Chinese' is not an exaggeration.
At the live launch event of Grok 4, the frequently appearing figure sitting next to Musk is Tony Wu, now a co-founder of xAI, who previously interned at Google DeepMind and OpenAI, did a postdoc at Stanford, and also worked at Google for a period.
Musk and Tony Wu (right) | Image source: xAI
This year, there is even a popular saying that gets joked about and forwarded in the tech circle from time to time: 'The AI war is Chinese people in the U.S. vs. Chinese people in China.'
03 Cutting jobs while poaching: 99% of the money flows to 1% of the people?
However, behind the glamorous sky-high contracts lies another group’s anxiety, as this poaching battle may only target the top 1% of the pyramid; what about the remaining 99%?
Even 'ordinary' senior AI engineers can earn between $1 million to $1.5 million annually, two to three times higher than traditional software positions. Data from the Levels.fyi platform shows that AI engineers at Meta at the E7 level can average annual packages nearing $1.54 million, a price that is considered upper tier even in Silicon Valley.
But for many Silicon Valley programmers, the rise of AI and the fierce talent battle among giants brings not only envy but also a real sense of crisis: on one side are giants like Meta, OpenAI, and Google competing for top AI scientists with signing bonuses in the tens of millions or even hundreds of millions, AI experts holding sky-high contracts, enjoying nine-figure salaries; on the other side are ordinary engineers worried about layoffs and being marginalized.
‘On one side, we see various LLM experts getting massive packages, and on the other side, ordinary engineers worry about being laid off every day.’ Someone posted this on a forum frequented by Silicon Valley coders, and there are many similar posts across various social networks in the Silicon Valley tech circle.
And giants are indeed 'laying off while poaching.' In recent years, Meta has laid off at least tens of thousands, especially employees not working on AI projects, implementing a 'rank and yank' system that this year was jokingly referred to as the 'squid factory' by the Silicon Valley Chinese coder community; Google is similarly optimizing and even initiated a 'voluntary departure compensation plan,' redirecting resources to AI projects; Amazon laid off more than 20,000 people last year and cut several corporate positions earlier this year, starting a restructuring of AWS-related departments in March.
In July 2025, Microsoft announced further layoffs of thousands, primarily focused on engineering positions, with hundreds of software engineering jobs cut in Silicon Valley alone, partly due to AI improving productivity.
Microsoft CEO Nadella | Image source: Microsoft
Microsoft CEO Nadella publicly stated in 2025 that 20% to 30% of the code within Microsoft is generated by AI. Similar situations have arisen in other companies; for example, Salesforce executives also stated that approximately 20% of the code within their company is generated by AI, with AI boosting engineering team productivity by over 30%, thus reducing the need for programmer recruitment.
Some Silicon Valley software engineers believe that as AI coding efficiency improves, the survival of ordinary software engineers is becoming 'increasingly difficult.' Some even think that currently, 99% of the money flows to 1% of top AI talent, and that there are not many AI positions available, with AI developed by programmers replacing many roles, which might ultimately lead to their own obsolescence.
The AI talent poaching war in Silicon Valley is not just a zero-sum game between giants. Whether it's AI talent, ordinary software engineers, or Silicon Valley giants, everyone is forced to accept this high liquidity and short-termism, as well as a reality:
A lot of money, even more money, is only flowing towards AI.
