Koray Kavukcuoglu has been appointed as the new head of Google's AI division, DeepMind, reporting directly to CEO Sundar Pichai.
He will oversee the development of the Gemini large language model, cutting-edge AI research, and the teams behind Gemini applications and developer tools. In the frontier AI arena, coding capabilities are becoming a critical battleground, and Google is striving to keep pace with OpenAI and Anthropic.
On May 19, 2026, at the Google I/O developer conference in Mountain View, California, a Gemini AI logo was on display. Google's parent company, Alphabet, unveiled a refreshed classic search box and a new AI-powered coding tool, marking the latest step in its multi-billion-dollar effort to expand its influence in the AI era.
The new DeepMind chief faces immense pressure to narrow the AI performance gap with OpenAI and Anthropic. In recent months, both OpenAI and Anthropic have released new models that have garnered praise for their power while also raising concerns. In contrast, Google, which acquired DeepMind in 2014, has not released a new generation of frontier models since early 2026.
Kavukcuoglu, who previously served as the AI division's Chief Technology Officer and Google's Chief AI Architect, has now officially taken on the challenge of leading the race to catch up in the frontier sector, with the goal of building the industry's top large language models. He was promoted to Senior Vice President, replacing DeepMind co-founder and CEO Demis Hassabis, who will transition to the role of Chairman of the organization.
Ben Wood, Chief Analyst at CCS Insight, said in an interview, "There's no doubt the goal is to close the gap with Anthropic and OpenAI in some areas." Wood added that Kavukcuoglu's promotion "sends a clear signal that Google plans to adjust its R&D focus, reducing academic-oriented projects and concentrating more on improving frontier model performance and enhancing the developer tool ecosystem."
Competitive Landscape in the Frontier Arena
After OpenAI led the industry following the public launch of ChatGPT in 2022, Google was initially lagging behind. The situation shifted in November 2025 when Google released Gemini 3. Analysts described the model as "pushing the frontier of the industry to a new level," with capabilities "far exceeding what the public generally expected from this generation of frontier large language models."
However, in 2026, OpenAI and Anthropic, the developer of Claude, regained the lead with new models. Anthropic's Mythos model achieved a significant breakthrough, while OpenAI's GPT-5.6 also received widespread acclaim. After releasing Gemini 3.1 Pro in February of this year, Google's subsequent products have failed to compete effectively in the frontier arena.
In response, Google formed an internal task force called "Code Strike" to bolster its coding capabilities, which are crucial for the development of artificial general intelligence (AGI). Malik Ahmed Khan, a senior equity analyst at Morningstar, analyzed that Google's lag behind Anthropic and OpenAI is "because Google's earlier focus was more on monetizable businesses like search and multimodal, causing it to miss the first killer application—code development. The two competitors have already far surpassed Google in coding capabilities."
A Google spokesperson responded, saying the claim that Google has abandoned its core goals and put the cart before the horse is inaccurate. The path to AGI requires simultaneous progress on multiple fronts, including agents, code, robotics, and world models. The spokesperson also highlighted DeepMind's achievements in fields such as robotics, video, on-device intelligence, public policy, and fundamental research.
DeepMind Under Kavukcuoglu
Kavukcuoglu is a DeepMind veteran, having joined the team in 2012, before Google's acquisition. With several key talents leaving the company, he needs to re-energize the team to compete head-on with Anthropic and OpenAI in the frontier arena.
In an official statement, Pichai said that Kavukcuoglu reports directly to him and is fully responsible for Gemini model development, frontier AI research, Gemini applications, and the developer team. Morningstar analyst Khan believes this appointment will make DeepMind's goals more focused, with optimizing large language models being a clear path forward. "After this adjustment, Google's position in the large language model race is likely better than before the adjustment."
"Hassabis was more inclined to explore paths to AGI beyond the scope of large language models," Khan said. During Hassabis's tenure as CEO of DeepMind, his focus encompassed multimodal, world models, and life sciences, such as Isomorphic Labs.
A source close to the DeepMind team told [source] last week that over the past year, Kavukcuoglu had gradually taken on more work from Hassabis, leading model development and making multiple stage appearances to launch important versions of Gemini.
Ray Wang, Principal Analyst at Constellation Research, commented that Kavukcuoglu's promotion means Google will prioritize execution over pure basic research. "We can expect a faster pace of model iteration, with more innovative experiments being implemented, all driven by the product roadmap."
In July, Bloomberg reported that Google delayed the release of Gemini 3.5 Pro to focus on optimizing performance in areas like code, and Alphabet's stock price fell on the news. Nick Patience, Head of AI at Futurum Group, said that Kavukcuoglu's primary task is to ensure the successful launch of Gemini 3.5 Pro. Secondly, he needs to stabilize the iteration cadence and prove this is not a one-time breakthrough. In addition to delivering a competitive frontier model, he "also needs to rebuild the code development and pre-training talent team to fill the gaps left by the departure of key personnel."
Before the acquisition, DeepMind was headquartered in London, and Hassabis remained based there during his tenure as CEO, with DeepMind's continued presence in the city boosting the UK's AI ecosystem. However, Kavukcuoglu, in his previous role, had already relocated from London to Google's Mountain View headquarters in California. Patience noted, "Google is quietly moving its core AI leadership from London." A Google spokesperson, however, stated that the company remains committed to its presence in London.
Outlook for Google's AI Strategy
On the day of the senior leadership change announcement, Alphabet's stock price dipped slightly, but over the past 12 months, the stock has gained approximately 76%. During last month's earnings call, Pichai mentioned that nearly 90 of the Fortune 100 companies are using Gemini Enterprise, highlighting Google's strength in selling AI services to cloud customers.
Sebastian Mallaby, author of "The Infinite Machine" (a biography of Hassabis), said in a program, "Although Google's models are temporarily behind and the company is eager to catch up, its progress in commercializing AI is very impressive." He pointed out that Google's AI technology is already integrated into Android phones and also powers Siri and Apple Intelligence on iPhones.
Mallaby argued that the departure of several high-profile talents should not be seen as a full-blown crisis for Google. "The team is huge; the Gemini-related team alone has thousands of people. There's no need to overstate the impact of a few core personnel leaving."
Brian Hopkins, Vice President and Principal Analyst at Forrester, proposed that placing model development, client applications, and developer teams under Kavukcuoglu's control, reporting directly to Pichai, could further amplify Google's AI competitive advantage. "Putting the large model, end-user applications, and developer ecosystem under the management of one executive is the organizational model of a mature product division, not a traditional lab structure. In this respect, Google is far more experienced than OpenAI and Anthropic."
Comments