"메타가 오픈AI 직원을 빼가려고 약 1억달러(약 1300억원)에 달하는 계약 보너스(sign-on bonus)를 제안했다."Sam Altman, OpenAI Chief Executive Officer (CEO), said, "Meta proposed a contract sign-on bonus of about $100 million (approximately 130 billion won) to lure away OpenAI employees."
As American Silicon Valley corporations are fiercely competing to secure artificial intelligence (AI) talent, the AI market is becoming similar to professional sports, where star players are acquired with high transfer fees and salaries. Recently, the value of 'S-level' AI talent soared to an annual salary of around 100 billion won, surpassing that of top players in the National Basketball Association (NBA). As the manpower for building and advancing large-scale AI models is limited, fierce competition for talent among Big Tech AI firms is expected to continue for the time being.
In contrast, Korea faces a serious outflow of AI talent, necessitating urgent measures to stem the brain drain.
◇ Big Tech becomes an 'AI talent black hole'
The company that ignited the intensified battle for AI talent this year is Meta, the parent company of Facebook. Mark Zuckerberg, the CEO of Meta, established the Meta Superintelligence Labs (MSL) for advanced intelligence research earlier this month. He has proposed an extraordinary compensation package exceeding $100 million to fill the lab with 50 of the industry's top AI talents, poaching core personnel from competitors.
According to major foreign news reports on the 31st, Meta has hired over 40 AI talents from Google DeepMind, Apple, OpenAI, and Anthropic alone this year. Most recently, Bowen Zhang, who led multimodal AI research at the Apple Foundation Model (AFM) team, moved to Meta. This means that Apple has lost four key researchers to Meta in the past month. Previously, Luming Fang, who oversaw the AFM team, along with core researchers Tom Gunter and Mark Rido, joined the Meta Superintelligence Labs. It is reported that Fang accepted an offer of over $200 million (approximately 280 billion won) to move to Meta.
Recently, Meta has hired more than 10 researchers from OpenAI, the developer of ChatGPT, in the past two months. According to a list of talents affiliated with the Meta Superintelligence Labs released by Didi Das, a partner at Menlo Ventures based in Silicon Valley, about 40% of the current 44 lab members are from OpenAI, 20% from Google DeepMind, and the rest from Anthropic and Apple. Among them, nearly half, including 21 people (47.7%), are nationals of China and graduated from prestigious Chinese universities such as Tsinghua University, Peking University, and Zhejiang University.
Zuckerberg noted, "We are focused on building the most elite AI team in the industry with the goal of achieving superintelligence." Superintelligence refers to AI that surpasses human intelligence, being a next-generation AI that exceeds 'artificial general intelligence (AGI)' with capabilities similar to those of humans.
Microsoft (MS) also recently hired over 20 AI experts from Google DeepMind. Amar Subramanya, an engineering vice president who led the development of the AI model Gemini Assistant during his 16 years at DeepMind, has joined MS as the vice president of AI organization (CVP). Last month, Adam Sadovsky, who was a senior engineer for 18 years at DeepMind, also moved to MS as a vice president. Earlier this month, Sonal Gupta, the engineering lead at Google DeepMind, also transitioned to the MS AI team. They are expected to enhance the AI model 'Copilot' and the search engine 'Bing' while developing new AI technologies within MS's AI organization led by Mustafa Suleyman.
The recruitment of AI talent by MS contrasts with the large-scale restructuring that the company is undergoing to cut expenses. Earlier this month, MS announced it would reduce its workforce by about 9,000 employees, approximately 4% of its global workforce, as a majority of jobs can be easily replaced by AI. Conversely, there is a pattern of aggressively recruiting top-tier AI talents with high salaries.
The compensation packages for the talents who moved to MS this time have not been disclosed, but given that the value of AI researchers and engineers has jumped at least 2 to 3 times in recent years, analysts suggest they will not be small.
According to recruitment company Harrison Clark, the annual salary packages offered by Big Tech to senior AI researchers have risen from the range of $400,000 to $900,000 (approximately 540 million won to 1.22 billion won) in 2022, the year ChatGPT was released, to between $500,000 and $1.5 million (approximately 680 million to 2.72 billion won) recently. The basic salary of a general senior software engineer without AI experience typically ranges from $180,000 to $220,000 (approximately 240 million to 300 million won), which falls below half of the AI engineer's level.
The average annual salary of a general AI engineer is between $3 million and $7 million (approximately 4.08 billion to 9.53 billion won), which has increased by about 50% compared to 2022. For top-tier AI talents, salaries exceed $10 million (approximately 1.36 billion won). The reason for the sky-high valuation of AI researchers is the extremely limited number of people capable of designing and advancing cutting-edge AI models like OpenAI's GPT model or Google's Gemini globally. While the demand for S-level talents is skyrocketing, the supply remains woefully inadequate.
The U.S. tech media TechCrunch stated, "The AI market has now entered a phase where securing top-level talent has become more important than data or computing power, and the heated competition for AI talents resembles that of the professional sports market."
◇ "Korea ranks fourth in AI talent outflow among OECD countries"
In contrast, Korea faces a serious outflow of AI talent. A report published last month by the Korea Chamber of Commerce and Industry (KCCI) Sustainability Growth Initiative (SGI) notes that last year, Korea's net outflow of AI talent was -0.36. If positive, it indicates an influx of talent, while negative implies an outflow, meaning Korea recorded the fourth highest outflow among the 38 member countries of the Organization for Economic Cooperation and Development (OECD), placing it at 35th.
Industry insiders explain that Korea's inadequate AI industry ecosystem compared to advanced countries, along with better compensation and research environments abroad, have led to the ongoing talent drain. A source from the domestic AI industry stated, "There is a salary disparity of over 10 times between high-level AI-related jobs in domestic and overseas corporations." Dr. Jong Hyoung-won, the only Korean researcher who participated in developing OpenAI's 'o1' model, also recently moved to the Meta Superintelligence Labs for a high salary.
According to a report from the Software Policy Research Institute (SPRi), 81.9% of the 2,354 domestic AI corporations are facing a shortage of personnel. Even AI experts nurtured domestically are leaving for better conditions abroad. The report further explains that domestic S-level talents tend to move to foreign corporations in the U.S. and Canada, while A-level talents are trending towards Naver and Kakao. In the manufacturing sector, even large companies find it challenging to procure AI talents.
Seo Seong-il, Vice Chairman of the Korea Software Industry Association (KOSA), stated, "More than 10,000 AI talents are expected to be lacking this year alone."
The Lee Jae-myung administration aims to strengthen sovereign AI and establish AI data centers, and secure graphics processing units (GPUs) with the objective of becoming an 'AI powerhouse.' However, there are concerns that without the necessary AI talents to support these initiatives, it will be difficult to achieve substantial results. Kim Myung-joo, Director of the AI Safety Institute, noted, "Competent domestic AI talents can only prefer Big Tech with better treatment and growth potential. While cultivating AI talent and preventing brain drain are essential, we must also consider ways to attract AI experts from China and India to our country."