Billions of dollars in Silicon Valley are competing for AI talent, and it is rumored that Zuckerberg has a "headhunter list" and is really anxious?

Silicon Valley erupted in the most intense AI talent war in history during the summer of 2025, with a secret list igniting a "billion-dollar salary" competition among human resources. (Background: Meta is frantically competing for AI talent, with Zuckerberg proclaiming "annual salaries exceeding $100 million"; Sam Altman sarcastically comments: can't buy the best employees.) (Additional context: OpenAI is rumored to launch an AI version of Office software, directly competing with Microsoft and Google.) This summer, the air in Silicon Valley smells of fresh cash, from Palo Alto to San Francisco, Highway 101 is being scorched by the heat wave, and tech companies keep their office air conditioning at maximum, yet it can't quell the anxiety of the HR departments, because Silicon Valley is entering a rare state of full "fire" for AI talent. "The List" ignites salary ceilings A secret list known as "The List" is circulating wildly in encrypted chat groups. This list has fewer than a hundred names, mostly from MIT, CMU, and UCB, specializing in speech recognition, machine learning, and generative AI. They are seen as strategic assets more scarce than advanced chips, and the market is quickly spiraling out of control: mid-level engineers start at an annual salary of $350,000, top researchers exceed $500,000, and signing bonuses are rumored to exceed $10 million. According to a report by the Economic Times in June, this wave of talent grabbing has pushed salary ceilings to unprecedented heights, truly scarily high. The poaching drama led by Zuckerberg Meta CEO Zuckerberg is no longer staying in the conference room; to build a foundation for a "superintelligence laboratory", he personally organizes "The List" and opens a "Recruiting Party" group chat to pump people. At a lakeside villa, researchers roast marshmallows while listening to him explain the AGI blueprint, and by the end of the gathering, they receive offers in the eight-figure range in USD. Meta's estimated capital expenditure in AI this year is $70 billion, and according to GetCoAI's report in June, the primary purpose of spending so much money is to "buy time, buy talent". Meta further invested $14 billion to acquire nearly half of Scale AI, inviting 28-year-old CEO Alexandr Wang to oversee research, making recruitment and investment seamless. Further reading: Meta frantically competes for AI talent, Zuckerberg proclaims "annual salaries exceeding $100 million"; Sam Altman sarcastically comments: can't buy the best employees. Tribal knowledge and chain-hopping The top researchers in AI are well acquainted with each other, with algorithmic details and hyperparameter insights referred to internally as "tribal knowledge", which is difficult to transfer through documentation and can only be brought into new companies through personnel movement. The poaching by these large companies exhibits a small group model of movement; for example, OpenAI's Sora team members Bill Peebles and Tim Brooks updated their LinkedIn profiles and then appeared together in Sunnyvale. OpenAI immediately restarted salary negotiations in an attempt to block the corner. Sam Altman publicly stated: "Using $100 million to poach our people? Unfortunately, you'll need to try harder." Poaching from The Wall Street Journal On July 1, it was revealed that Meta managed to poach eight OpenAI employees within two weeks, which for a startup with limited funding means that the annual salary of a PhD-level researcher could consume the entire funding round. This is the showtime for big companies, while startups can only watch enviously or wait until they are poached as well. Is research freedom more enticing than cash? While cash is flying everywhere, many researchers value research freedom and the allure of "unsolved problems" even more. For instance, Berkeley professor Alexei Efros, regarded as a conscience of the industry, often reminds students that the goal of work is to unlock "cool, important, yet unanswered mysteries". This saying has become a code of conduct in Silicon Valley: only companies with massive computing power and data can provide the stage needed, attracting a certain type of researcher. For example, speech recognition was once thought to be an AI dead end, but has returned to industry focus thanks to deep learning, serving as the best example. For many researchers, the ability to operate thousands of H100s and quickly validate hypotheses is more attractive than any stock options. This is understandable; once humanity harnesses the power of nuclear weapons, it cannot return to fighting with a spear. Starting from the summer of 2025, the worth of PhD-level researchers has been pushed towards billion-dollar targets. Meta is spending money, Google is banding together, and OpenAI is countering, as the tension in Silicon Valley continues to rise amidst the late-night lights. Money can buy short-term victories, but what truly shapes the future of AI are those minds willing to challenge unsolved problems; the era of superintelligence is quietly being sketched out amid the repeated job-hopping.