On the same week that Google reported its cloud business growing at a blistering pace, it announced that Jeff Dean, the man who built much of the technical foundation of modern Google, was leaving after 27 years.
At first glance these two facts do not fit together.
How can a company be triumphing in artificial intelligence, its commercial engine roaring, while the very people who created that technology head for the exits?
The answer is that success and departure are not opposites here. They are, in a sense, connected, and understanding why explains a great deal about where the AI industry is heading.
Commercial triumph
Start with the good news, because it is substantial.
Google's cloud division, which rents out computing power and AI tools to other businesses, reported revenue up 82% year on year, an extraordinary rate of growth for a business of its size.
The company has managed to turn its AI research into products that customers are actually paying for, with its Gemini Enterprise software now used by around 90% of the Fortune 100, the largest companies in America.
It is even selling its own custom AI chips, known as TPUs, to outside buyers including Anthropic, one of its direct rivals in building AI models.
This is what winning looks like in the current market, converting years of expensive research into revenue, and Google is doing it convincingly.
Talent exodus
Now the other side. Dean is not leaving alone, and that is the part that should give Google pause.
He is departing with three other senior figures, Sanjay Ghemawat, Oriol Vinyals and Quoc Le, to start a company called Discovery Loop, aimed at using AI to automate scientific research itself.
These are not ordinary employees, but among the architects of the systems that made Google's AI possible in the first place.
Their exit caps a remarkable run of departures, and by one count all eight authors of a landmark 2017 research paper, the one that introduced the "transformer" design underpinning virtually every modern AI system including Google's own, have now left the company.
In June alone, one of Gemini's co-creators left for OpenAI and a Nobel prize-winning researcher left for Anthropic, and Alphabet's share price fell sharply on the news, slipping again this week.
Why the two things go together
Here is the resolution to the apparent paradox. The skills that invent a technology and the machinery that sells it are different things, and a company optimised for one is often a frustrating home for the other.
A business generating 82% cloud growth is, by necessity, focused on products, deadlines, customers and the quarterly earnings that public companies live by.
That is precisely the environment a certain kind of researcher wants to escape, and Dean said as much, explaining that operating outside a public company gives his team room to make choices that are not always in a corporation's purest financial interest.
In other words, the same commercial machine that produces Google's success can make it a less appealing place for the people who prefer open-ended discovery to shipping features on a schedule.
Commercialisation does not merely coincide with the talent drain, it helps cause it.
New team, same story
The leadership changes announced alongside Dean's exit point in the same direction.
Demis Hassabis, the Nobel laureate who runs Google's DeepMind AI lab, is stepping back from day-to-day management to become its chairman and Alphabet's chief scientist, a role focused on the big picture rather than daily operations.
His deputy, Koray Kavukcuoglu, takes over the practical job of running DeepMind and shipping the next version of Gemini.
That is a classic move for a company shifting from a research-led culture to an execution-led one, putting an operator in charge of turning ideas into products on time.
It suits the phase Google is now in, but it is not the culture that keeps restless inventors in their seats.
What it means
The important thing for the average reader is to see that these stories are about different timeframes.
The 82% growth reflects the past, the payoff from research done years ago, now being harvested.
The departures are about the future, a risk that the people best able to invent the next breakthrough are choosing to do it elsewhere, often with Google's own money, since Alphabet is backing Dean's new venture as an investor.
Google can be both a present-day champion and a company quietly losing its edge for the next round, and for now it is both at once.
Whether today's profits or tomorrow's brain drain matters more is the question its investors are now nervously trying to price.
