Let us know about free updates
Simply sign up for the technology sector Myft Digest and it will be delivered directly to your inbox.
The author is former editor-in-chief of Wired Magazine and writes FuturePolis, the newsletter for Future of Democracy
Even by Silicon Valley’s historically rare standards, Big AI is spending a stratospheric amount of talent this year. META is investing $15 billion in Scale AI, a data label startup that claims just 900 employees. Scale’s 28-year-old CEO Alexandr Wang will take on the job at a new Metallab dedicated to creating AI’s “Superintelligence.” His cash and stock in the transaction are reported to be worth around $5 billion, making him one of the most expensive so-called “acquisters” on record.
Meta also reportedly offers a $100 million sign-on bonus to lure researchers from other artificial intelligence companies into the lab. Meanwhile, Openai paid $6.4 billion to IO, a boutique design company led by former Apple’s top designer Jony Ive. Also, a bid war from rivals trying to hire top researchers has led to employees who have already reached eight figures paying up to $2 million in retention bonuses to employees who have already reached eight figures.
What lies behind this gold rush? It’s not a lack of talent, itself. The San Francisco Bay Area is being attacked by unemployed software engineers. This is the result of the industry cutting jobs and adopting AI coding tools after the pandemic.
Rather, the eye-catching person is a marker of how difficult it is for the biggest AI companies to build a “moat” or competitive advantage that is impossible to attack. Their models are taking away the top spot in performance, but they are a crude and inexpensive alternative from rivals like China’s Deepseek Nip on their heels. The data centers and the chips that fill them are products, although very expensive. That leaves two areas where businesses hope to steal March: data and talent.
Scale AI is both talent and data play. The company’s main business is to provide high quality annotated data to train AI models. Now that large AI companies have cut most of the internet, scale labeling work can help them improve the quality of their models. Meta’s Mark Zuckerberg must be desperate to continue his company as a serious player in AI after the latest, massive model release, Lama 4, was dominated by.
But war for talent is about recognition, not just production. The ability of startups to attract investors and the ability to maintain the stock price of publicly traded companies is both aided by the topics produced by the minds of a few superstars.
Compared to the expected revenue and team size, scale is one of the most expensive major technology acquisitions to date. But that’s not just one time. In 2014, Facebook acquired WhatsApp for $21.8 billion when there were only 55 people from the messaging company. Founders Brian Acton and Jan Calm both joined Facebook as part of the deal.
Still, the amount offered to AI is rare. There are three sources for all these dollars. The first is the relentless pursuit of profits from large AI and chip companies, bet on the US government’s resolve to maintain America’s lead in AI surrounding China.
The second is the fierce pace of advancements in AI tools and the rush by other industries to sprinkle all of AI dust on top for fear of falling behind. While many of these investments have not yet brought about productivity gains, FOMO is a powerful force.
The third factor is that for better or worse, large AIs have been trapped in artificial general information or AGI competition. This is a conceptual point where AI reaches and surpasses human capabilities. In reality, some of the world’s most legendary AI experts, including Meta’s own Yang Lekun, argue that large-scale language models do not do tricks and require new approaches to study. The AGI itself may be Mirage. The definition of that may be very different, and that what the future holds is not human-like AI, but many different, highly capable, more specialized kinds.
This is why races for this talent require more than just cash. There is also culture and mission. Both are artificial and safe tensions created by former Openai employees, focusing on creating, for example, “safe” AI. Humanity, reportedly being given more autonomy to researchers, is suitable for retention.
Meanwhile, Openai has lost many of its best people in recent years after Rifts with top leadership. In Meta, researchers cite blue ski research neglect in their reasons for retirement. This year, it lost the fair’s director, Joel Pineau.
These wars of talent show no signs of slowing down. The AI chief has staked his reputation as being the first person in AGI. As long as the dreams of greed, fear and super intelligence are maintained, enormous wealth for the best talent continues to flow.
