Google is entering a new phase in the global battle for artificial intelligence. As competition with OpenAI, Anthropic and the other major players in the sector intensifies, the American group is reorganizing its AI leadership while seeing some researchers and engineers who helped lay the scientific foundations of today’s technologies depart.
Behind these departures, the financial question does not tell the whole story. The freedom to research, the speed of decision-making, and the possibility of owning a significant stake in young companies also become decisive arguments for attracting the sector’s best specialists.
For Google, the stakes are now double: to continue accelerating the commercial development of Gemini while retaining the scientists capable of preparing the next generation of technologies.
Google reorganizes DeepMind leadership
This transformation begins with a new distribution of responsibilities within Google DeepMind, its main artificial intelligence lab.
Demis Hassabis, co-founder and former CEO of DeepMind, shifts to a role more focused on strategic research. He becomes, according to the information provided, Chief Scientist of Alphabet and Chairman of the Board of DeepMind.
This shift should allow him to devote more time to long-term scientific projects, notably artificial general intelligence, or AGI, which remains one of the main objectives pursued by the world’s leading AI laboratories.
The daily operational management and the development of the Gemini models are entrusted to Koray Kavukcuoglu, an experienced DeepMind researcher who has worked notably in reinforcement learning and speech processing.
Google thus seeks to further separate scientific leadership from operational management in order to speed up the transition of lab advances into products that can be deployed at scale.
When researchers want to go faster
This reorganization takes place in an environment where speed has become a major driver of competitiveness.
Large tech companies have substantial budgets, powerful data centers, and thousands of engineers. But their size can also slow certain decisions.
In an organization the size of Google, a scientific innovation often has to pass several stages before becoming a product: technical validation, safety, compliance, integration into the existing ecosystem, and commercial trade-offs.
A young company specializing in artificial intelligence can operate differently. A small team can quickly decide to develop a new idea, launch a prototype, then improve it based on user feedback.
This difference in pace has become particularly visible since the emergence of ChatGPT. OpenAI has shown that a model arising from research can be transformed very quickly into a product used by hundreds of millions of people.
For researchers, the ability to experiment and then rapidly launch their work can be as attractive as a high salary.
Jeff Dean and the weight of AI’s big names
Among the departures noted is Jeff Dean, one of the most influential engineers in Google’s history.
He is among the leading figures of Google Brain, the team that played a central role in the development of deep learning technologies now essential to many modern AI systems.
Jeff Dean also contributed to building computing infrastructures enabling Google to train very large models as well as to the development of technologies related to AI-specialized processors.
According to the information provided, he is leaving Google to found a new company called Discovery Loop, whose objective would be to use artificial intelligence to accelerate scientific discoveries and engineering work.
This departure would fit into a broader movement: several AI researchers have left Google in recent years to join competitors, notably OpenAI and Anthropic, or to start their own companies.
Money is no longer enough to retain the best researchers
In the current talent competition, compensation can reach extremely high levels. But industry specialists emphasize that money is no longer the sole criterion.
Startups can offer researchers three particularly attractive advantages: greater scientific autonomy, shorter decision-making chains, and meaningful equity stakes.
This latter dimension can profoundly change the equation.
A researcher working for a large company generally enjoys a high salary and substantial resources. In a fast-growing startup, they can also become a shareholder in a company whose value could rise very sharply.
For the most sought-after profiles, the choice is no longer limited to comparing two salaries. It also means choosing between working in a gigantic organization or directly participating in building a new company.
Gemini becomes central to Google’s strategy
For Google, this talent war comes as Gemini takes an increasingly central place in its strategy.
The group aims to integrate its AI into virtually every component of its ecosystem: Search, Android, Google Workspace, Google Cloud, and many other services.
Google thus has a significant advantage. Few companies can deploy a new technology to such a large user and business base.
The group also benefits from massive computing infrastructure, specialized processors, extensive research capabilities, and a long track record in algorithm development.
But these hardware resources are no longer enough.
The performance of future models also depends on the ability to recruit and especially to retain the scientists who imagine new architectures, training methods, and applications.
AI’s new rare resource
The departures of prestigious researchers do not mean that Google loses its ability to compete with its rivals.
The company still maintains one of the world’s largest pools of AI expertise, as well as infrastructures and financial resources that few players can match.
However, these movements reveal a profound transformation of the industry.
For several years, competition in AI has primarily been measured by compute power, available data, and billions of dollars invested. These factors remain essential, but another resource is becoming increasingly scarce: researchers capable of producing the next technological breakthroughs.
Google is thus trying to solve a complex equation. The company must operate quickly enough to compete with specialized startups while preserving the advantages of a global, integrated group.
DeepMind’s reorganization and the weight now given to Gemini illustrate this search for balance.
Because in the new AI battle, the winner may not be the one who owns the most powerful model or the largest number of data centers. It could also be the one who manages to create the environment in which the best scientists want to stay, experiment, and build the next generation of AI.
