CNBC: Kavukcuoglu inherits DeepMind’s catch-up race — OpenAI and Anthropic stay the yardstick
Koray Kavukcuoglu did not inherit a victory lap. He inherited a scoreboard. In a August 12 follow, CNBC frames DeepMind’s new boss as the executive charged with closing the frontier and coding gap with OpenAI and Anthropic while Demis Hassabis moves to Alphabet-level AGI strategy. The org chart story already shipped. This is the catch-up story.
What CNBC says the mandate is
CNBC’s through-line is blunt. Kavukcuoglu, previously DeepMind CTO and Google’s chief AI architect, becomes head of DeepMind reporting to Sundar Pichai. He oversees Gemini model development, frontier AI research, and the Gemini app and developer teams. Analysts quoted in the piece treat that remit as a product-group seat, not a lab hobby.
CCS Insight’s Ben Wood told CNBC the goal will “undoubtedly” be closing gaps with Anthropic and OpenAI in some areas, and that the promotion feels designed to shift focus from academic projects toward frontier performance and developer tooling. Morningstar’s Malik Ahmed Khan said LLM improvements are now the clear route forward, and that Google looks better positioned in the LLM race with the change than without it. That is the mandate CNBC is selling: ship competitive frontier models, faster.
Coding and frontier pressure points
The competitive bruise is specific. CNBC notes Google has not unveiled a frontier model since early 2026, while OpenAI’s GPT-5.6 and Anthropic’s Mythos drew the “advanced capabilities” headlines. Gemini 3 in late 2025 briefly reset expectations. Gemini 3.1 Pro in February did not keep Google at the tip of the spear.
Coding is the pressure point analysts keep naming. Khan told CNBC Google lost the first clear killer use case because it prioritized monetizable areas such as Search and multimodal work, leaving Anthropic and OpenAI “miles ahead” on coding. Alphabet assembled an internal “Code Strike” team to claw back. Bloomberg’s July report that Gemini 3.5 Pro slipped to improve coding performance still hangs over the cadence.
Hassabis vs ops split, restated
CNBC restates the split for a business audience. Hassabis becomes chair; Kavukcuoglu takes day-to-day. Khan’s read: Hassabis was more interested in AGI beyond LLMs, with multimodal, world models, health, and Isomorphic Labs in the mix. Ray Wang of Constellation Research called the promotion a bet on “execution over deep research,” expecting faster releases, more experimentation, and a product roadmap emphasis.
Geography is part of the signal. DeepMind’s London roots and Hassabis’s UK presence helped build a U.K. AI ecosystem. Kavukcuoglu had already moved to Mountain View under his prior role. Futurum’s Nick Patience told CNBC Google is “quietly consolidating its AI leadership out of London.” Google said it remains committed to the capital. The practical meaning for the catch-up race: model and product authority sits closer to Pichai’s California operating rhythm.
Signals to watch in the next Gemini cycle
Patience’s near-term checklist is the cleanest watchlist in the piece. First, ship Gemini 3.5 Pro if possible. Second, prove it was not a one-off with a predictable release cadence. Third, rebuild coding and pretraining expertise after high-profile exits. Forrester’s Brian Hopkins noted the upside of putting models, the app, and developer teams under one executive: that is how you organize a product group, and Google historically does that better than OpenAI or Anthropic.
Monetization is Alphabet’s counter-argument. Nearly 90% of Fortune 100 companies use Gemini Enterprise per Pichai’s last earnings call, and Sebastian Mallaby told CNBC’s Squawk Box that painting a “general crisis” from famous departures over-indexes on a few names inside a multi-thousand-person Gemini effort. Analysis: revenue can be fine while the frontier narrative still hurts talent, partners, and next-cycle leverage. Kavukcuoglu’s job is to make the next Gemini drop end that argument without waiting for another analyst roundtable.
