The Google AI leadership reshuffle reorganizes Gemini engineering units to streamline research and counter aggressive rival moves.
Google co-founder Sergey Brin has urged core AI teams to focus entirely on developing the Gemini model as parent company Alphabet works to close the gap with industry rivals.
Facing pressure as rival Anthropic unveiled a preview of its powerful Claude Mythos model, Brin urged hundreds of Google DeepMind employees during an April town hall to accelerate their pace of development.
Although Google’s Gemini model briefly surged ahead of rivals last November, subsequent upgrades from Anthropic and OpenAI quickly pushed Alphabet back into catch-up mode. By August, Google was forced to delay its flagship Gemini release by two months after internal benchmark testing revealed its performance particularly in key areas like coding was still lagging behind competitors.
On August 5, tech giant Google announced a major leadership overhaul at Google DeepMind. Division head Demis Hassabis stepped aside from daily operations to become chair, handing executive leadership to his deputy, Koray Kavukcuoglu. Simultaneously, Gemini’s two original technical co-leads departed to launch an independent startup.
Despite speculation that Google might step back from the AI model race, interviews with insider sources reveal that Alphabet’s recent executive reshuffle and commercialization strategy are designed to double down on achieving model supremacy.
These changes reflect a continued erosion of DeepMind’s independence—an autonomy Google has steadily reduced since acquiring the London-based AI lab in 2014 to drive commercial software and revenue generation. As part of the recent corporate shakeup, Alphabet disclosed in an August 6 all-hands meeting that several DeepMind teams are being transferred directly into core corporate Google.
Details regarding the internal meetings and Sergey Brin’s remarks are reported here for the first time. Other media outlets previously reported on pending organizational restructurings and Brin’s personal coding push.
Google declined to comment on Sergey Brin’s recent involvement, DeepMind’s organizational restructuring, or employee concerns raised regarding the changes.
Under Demis Hassabis, Google DeepMind maintained a broader research mandate than its rivals while simultaneously developing the core Gemini models critical to Alphabet’s business. However, past reporting revealed that Hassabis had previously resisted certain commercial initiatives that could have boosted short-term revenue or competitive standing—a claim his allies dispute, arguing he never sacrificed business priorities for pure science.
In 2025, CEO Sundar Pichai named then-DeepMind CTO Koray Kavukcuoglu as Google’s chief AI architect to drive Gemini’s integration across company products. Insiders report that as Kavukcuoglu emerged as the central architect of Gemini’s roadmap, other senior leaders within DeepMind saw their influence decline.
Investor uncertainty surrounding the announcement sent Alphabet shares tumbling 4% on the day of the news. In response, Hassabis and Kavukcuoglu held a unit-wide all-hands meeting last Thursday to address widespread staff anxiety regarding the lab’s direction particularly since most employees had learned of the leadership overhaul only after it was publicly announced.
According to three sources with direct knowledge of the internal meeting, Google leaders reassured staff that day-to-day operations at DeepMind would remain unchanged. Sources added that Koray Kavukcuoglu will continue leading Gemini development, an operational role five sources noted he had already assumed more than a year prior.
According to three sources familiar with the meeting, Demis Hassabis delivered a presentation on how AI could accelerate scientific breakthroughs and reshape society long-term priorities that will form the core focus of his new role as chair.
Despite executive assurances, the all-hands meeting sparked fresh concerns that DeepMind’s independence is being swallowed by core Google. Hassabis and Kavukcuoglu confirmed that several nontechnical teams are being transferred into the corporate reporting structure. For a unit that once encouraged exploratory side projects, the directive was clear: refocus entirely on reclaiming model dominance.
Alphabet executives, including CEO Sundar Pichai, maintain that Google remains at the industry frontier in specialized capabilities like video generation, while affirming their commitment to closing performance gaps in lagging areas despite recent delays.
Meanwhile, senior leaders within Google Cloud expect Koray Kavukcuoglu’s appointment to reduce internal friction with DeepMind over how the company allocates its limited supply of custom AI chips, known as Tensor Processing Units (TPUs).
See also: Meta and 29 States Head to Court in Landmark Youth Social Media Addiction Trial
The Google reorganization occurs amid a renewed push by co-founder Sergey Brin to rally engineering teams mirroring his hands-on intervention following ChatGPT’s launch in late 2022, which initially put Google on high alert.
Despite having no official executive title, 52-year-old co-founder Sergey Brin has been actively shaping model training efforts. Sources familiar with his involvement note that Brin has spent several months urging key technical staff to push harder to match the capabilities of frontier competitors.
According to the source, Brin has leveraged his implicit authority as co-founder to redirect engineering resources toward frontier initiatives like recursive self-improvement—the threshold where AI systems can autonomously refine and upgrade themselves without human intervention.
Google declined to make Sergey Brin available for an interview. Meanwhile, sources indicated that Demis Hassabis had already scaled back his day-to-day oversight of Gemini, transferring management duties to key deputies shortly after launching the flagship AI model in 2023.
Although Google temporarily caught up to frontier competitors with the launch of Gemini 3 last November, it quickly lost its edge. Four sources attributed the company’s renewed laggard status to deep-seated structural challenges the same issues that caused it to fall behind initially.
According to internal sources, compute capacity constraints and leadership friction over project priorities starved critical areas such as coding capabilities of needed resources. Furthermore, sources noted that Google’s bureaucratic culture slowed model release cycles relative to nimbler competitors.
While Google declined to comment on these specific claims, CEO Sundar Pichai told Wall Street analysts last month that the company is continuing to aggressively scale its infrastructure. The expansion aims to relieve TPU bottlenecks for frontier model training while supporting both consumer and enterprise products.
Against this backdrop, Koray Kavukcuoglu has steadily consolidated his influence. Following his promotion to chief AI architect last year a position reporting directly to Sundar Pichai he relocated from London to Google’s headquarters in Mountain View.
According to a source with direct knowledge, several London-based leaders saw their influence over Gemini’s technical direction diminish following Hassabis’s step back from day-to-day operations and Kavukcuoglu’s relocation to Google headquarters.
Outside of research circles, Kavukcuoglu established a reputation across Google as a pragmatist attuned to product commercialization. According to a source with direct knowledge, even prior to his physical relocation, Kavukcuoglu rather than Hassabis served as the primary bridge between DeepMind and Google Cloud, Alphabet’s central AI revenue engine.
A Google spokesperson confirmed to Reuters last week that Kavukcuoglu will now hold ultimate decision-making authority for major DeepMind initiatives.
High-profile leaders beyond Kavukcuoglu including Gemini’s original technical co-leads Jeff Dean and Oriol Vinyals departed during the recent leadership shakeup. Three sources noted that teams reporting to these departing figures previously enjoyed rare autonomy to pursue exploratory research separate from core Gemini model development.
Following their departures, Google provided no immediate clarity on the future of those affected teams or projects. The silence raised concerns among staff that DeepMind’s traditional defense against purely commercial pressures is continuing to erode.
