Behind Google's AI Leadership Change: Brin Pushes Gemini to Catch Up with Anthropic and OpenAI

Alphabet (GOOGL-US), Google's parent company, has recently launched a major AI organizational restructuring within its subsidiary Google DeepMind. While appearing as a high-level executive reshuffle, the move reflects Google's growing pressure to catch up in the generative AI race against rivals like Anthropic and OpenAI.

According to Reuters, citing seven individuals familiar with Gemini's development, Google co-founder Sergey Brin has been consistently urging the AI team over recent months to fully commit to Gemini, accelerate model development, and regain a leading competitive position against Anthropic and OpenAI.

On August 5, Google DeepMind CEO Demis Hassabis announced he would step down from day-to-day management, transitioning to Chairman of Google DeepMind and Chief Scientist at Alphabet. Operational leadership will now be taken over by his long-time deputy and former DeepMind CTO, Koray Kavukcuoglu. Kavukcuoglu will report directly to Google CEO Sundar Pichai and oversee the Gemini model, frontier AI research, and Gemini applications and developer teams.

This restructuring coincides with the departure of several core AI talents. Key figures such as Jeff Dean, a longtime leader in Google AI, and Oriol Vinyals, one of the original technical leads for Gemini, have left Google to join new ventures. After 27 years at Google, Dean departed to launch Discovery Loop, a startup focused on using AI to automate scientific and engineering research. Vinyals, Quoc Le, and Sanjay Ghemawat have also joined the entrepreneurial team.

Reuters notes that this adjustment is not merely a personnel change but reflects Alphabet’s attempt to reconfigure AI R&D authority—bringing Gemini, previously operating under DeepMind’s relatively independent research framework, more tightly into Google’s overall business strategy. During an all-hands meeting on August 6, management revealed that certain teams would be transferred from DeepMind into Google’s corporate structure, further weakening DeepMind’s historically autonomous operational model.

This shift aligns closely with Google’s recent push to directly monetize AI. Since acquiring DeepMind in 2014, Google has prioritized foundational research and long-term AI breakthroughs. However, with the emergence of ChatGPT and intensifying competition in generative AI, Google increasingly demands that Gemini directly serve search, cloud, enterprise software, and consumer products.

In 2025, Google CEO Pichai further appointed Kavukcuoglu as Google’s Chief AI Architect, adding to his existing role as DeepMind CTO the responsibility of embedding Gemini across Google’s product suite. Multiple insiders indicate that since then, Kavukcuoglu’s influence over Gemini’s direction has steadily increased, while the decision-making power of certain DeepMind executives has relatively diminished.

Kavukcuoglu has also become a crucial bridge between DeepMind and Google Cloud. As Google Cloud has emerged as a key engine for Alphabet’s AI commercialization, Kavukcuoglu, compared to Hassabis, has been more directly involved in coordinating between the two organizations. His appointment to lead DeepMind is seen as positioning him to better align AI model development with Google’s commercial needs.

This organizational shift occurs as Gemini faces mounting competitive pressure from Anthropic and OpenAI. Gemini briefly outperformed rivals in November last year, but after Anthropic and OpenAI released updated models, Google has once again fallen behind. According to sources, internal Google testing shows that the new flagship version of Gemini still lags behind competitors in certain capabilities such as programming, resulting in a roughly two-month delay in its release.

As a result, Google co-founder Brin has re-engaged more deeply in AI model training. Although 52-year-old Brin currently holds no formal management role at Google, insiders say he has repeatedly emphasized to key AI personnel the need to catch up with the AI frontier and has used his co-founder influence to drive resource allocation toward specific research areas.

One area receiving Brin’s attention is 'recursive self-improvement,' where AI systems can enhance their own capabilities with minimal human intervention. Reuters reported that Brin urged Google DeepMind to accelerate progress during an April town hall meeting attended by hundreds of employees.

This is not the first time Brin has intervened during a critical phase of the AI race. After ChatGPT debuted in late 2022 and shocked the tech industry, Brin also re-engaged with Google’s AI efforts. Now, as Anthropic and OpenAI continue rolling out more powerful models, he is again applying informal pressure internally—highlighting the rising strategic importance of Gemini within Alphabet.

Internal reactions to this restructuring at Google have been mixed. On the day Alphabet announced the DeepMind leadership overhaul—August 5—the company’s stock price took a hit, dropping about 4%. Employees also worry that DeepMind’s historically free culture of pursuing non-Gemini research may gradually disappear.

During a DeepMind-wide meeting on August 6, Hassabis and Kavukcuoglu attempted to reassure staff, stating that actual daily work would not undergo significant changes. Insiders note that Kavukcuoglu will remain focused on Gemini, while Hassabis will dedicate more time to long-term AI research, scientific advancement, and societal impact issues.

Nonetheless, the meeting confirmed that certain non-technical teams would be moved out of DeepMind and into Google’s corporate structure, making employees feel even more acutely that DeepMind’s autonomy is diminishing. For Google, this signifies a greater concentration of AI research resources on Gemini and projects capable of delivering direct commercial value.

Google’s current AI business performance is actually quite strong. Alphabet reported second-quarter 2026 revenue of $119.8 billion, a 24% year-over-year increase, with operating income up 30% and operating margin expanding two percentage points to 34%. Google Services revenue rose 15% to $94.5 billion, while Google Cloud revenue surged 82% to $24.8 billion, demonstrating that AI infrastructure and enterprise AI demand are rapidly driving cloud growth.

Gemini’s usage scale is also expanding quickly. Pichai stated in the Q2 earnings call that the Gemini app now has 950 million monthly active users, with daily active users tripling over the past year. Google also reported that nearly 90% of Fortune 100 companies have adopted Gemini Enterprise, and the Gemini model processes approximately 22 billion tokens per minute via API.

Demand for AI on Google Cloud has become a key backdrop for this organizational restructuring. Alphabet reported in Q2 that Cloud revenue reached $24.8 billion, up 82% year-over-year, with cloud backlog totaling $514 billion. The company noted that AI infrastructure and enterprise AI solutions are the primary drivers of Cloud’s rapid growth.

However, Google still faces limitations in AI computing resources. Pichai stated in the Q2 earnings call that the company will continue expanding infrastructure investments to address supply constraints in Tensor Processing Units (TPUs), supporting both AI model training and demand from consumer and enterprise products. Alphabet also indicated that demand for AI products from enterprises and consumers has clearly exceeded existing supply capacity.

TPU resource allocation has even become an issue between different departments within Google. Reuters cited sources indicating that Google Cloud executives hope that with Kavukcuoglu taking over DeepMind, tensions over limited TPU resources between DeepMind and Cloud can be eased.