Google Deepmind might lose its independence, and founder Demis Hassabis could quickly depart the corporate for good.
In accordance with a Guardian report, a former Google supervisor believes the “period of DeepMind as an unbiased actor is over,” and a present DeepMind worker now calls the lab “simply one other subdivision” of Google. AI researcher Koray Kavukcuoglu will take over day-to-day operations from the U.S., however he will not carry the CEO title. That alone backs up the declare: DeepMind not has its personal chief government.
As for Hassabis, Pathfounders reports that he needed to depart similtaneously Jeff Dean, however Google feared a good steeper inventory drop than the one which already hit when Dean’s departure and Hassabis’s function change had been introduced. As a substitute, Hassabis was “promoted” to chairman to clean the best way for an orderly exit. He plans to spend extra time on his scientific work and the pharmaceutical startup Isomorphic Labs and will depart DeepMind for good within the coming months.
With Hassabis sidelined, all Gemini-related AI improvement is shifting to the Bay Space, and Google co-founder Sergey Brin is anticipated to tackle much more management. Brin had already come out of retirement during Google’s first “red alert” over ChatGPT and jumped again into lively improvement. Shortly after, Google Brain and DeepMind merged into “Google DeepMind.” That is additionally how the Gemini models got their name, because the “twin” groups from each organizations had been imagined to work collectively on a joint undertaking. The tensions return years: even then, reviews surfaced about frustration with the restructuring, especially from Hassabis himself.
Is Google the following IBM or the following Westinghouse?
Google is as soon as once more in peril of falling completely behind the leaders. There are two methods to learn what’s occurring. The primary comes from SemiAnalysis in its newsletter “Gemini is Cooked but GCP is Cooking.” Dylan Patel and his colleagues see the shakeup as an indication that Google is dropping the race for high AI fashions.
Their argument is that DeepMind is not a frontier lab. The departures stem from years of overly cautious compute allocation and a bureaucratic, risk-averse tradition. Google had superior AI methods lengthy earlier than ChatGPT however did not get them into customers’ palms quick sufficient.
SemiAnalysis compares this to IBM retreating from the PC enterprise into mainframe consulting, or Intel giving up daring investments to guard its legacy enterprise. Each turned out to be errors.
One other piece of the critique is that Google prioritizes short-term cloud earnings by promoting compute to 3rd events as a substitute of giving its personal groups the assets to construct a lead in AI purposes. Gemini’s annualized income stood at $12 billion within the second quarter of 2026. Google Cloud, then again, is projected to generate over $73 billion in AI infrastructure income and a further $120 billion in TPU gross sales by the top of 2027. That is the place the cash is, not less than for now.
The second perspective comes from Web 2.0 inventor Tim O’Reilly in his Substack newsletter “Asimov’s Addendum.” O’Reilly argues that Google is intentionally operating a special race, specializing in infrastructure moderately than mannequin management. In main tech shifts, he says, it isn’t the inventors who win however the ones who unfold the know-how.
His go-to instance is Thomas Edison, the pioneer of electrical energy, who was overtaken by George Westinghouse and his Westinghouse Electrical Firm. Edison constructed the sunshine bulb and the primary energy plant, however Westinghouse gained through the use of alternating present to construct an influence grid that introduced electrical energy to all of society. Google’s latest Flash models fit this reading, since they’re optimized for effectivity whereas staying good sufficient to be helpful.
Google desires to guide in AI however cannot sustain
The rumor mill tells a special story, although. Based mostly on every little thing we have heard in latest months, Google did attempt to prepare frontier fashions, however the workforce could not catch the leaders, especially in coding performance. Gemini 3.1 Professional remains to be solely accessible in preview mode, and three.5 Professional, which was announced in May for a June release, has been roughly shelved. Each level to critical issues in AI improvement.
The latest restructuring could not simply replicate a strategic pivot towards infrastructure. It could merely be the results of the earlier workforce’s failure to ship. While you maintain all of the playing cards and nonetheless fall behind, frustration builds quick, and even the largest names within the trade aren’t untouchable anymore.
My learn is that Google finally desires each: the infrastructure and the lead in AI. They’re going to push more durable than ever to get there, with no regard for what got here earlier than. If there’s one takeaway from the previous few days, it is that.
AI Information With out the Hype – Curated by People
Subscribe to THE DECODER for ad-free studying, a weekly AI e-newsletter, our unique “AI Radar” frontier report six occasions a yr, full archive entry, and entry to our remark part.
