As recently as 2–3 years ago, Google was called a "lagging dinosaur" in the Generative AI race. When OpenAI unveiled ChatGPT in late 2022, a sense of crisis swept both the market and Google's ranks that the company's business model, built on search and advertising, could be shaken to its core. Some even said "the era of Google is over," and the hastily released chatbot Bard failed to regain trust amid controversy over inaccurate answers.
But the landscape began to shift in the second half of last year. Cracks appeared in ChatGPT's solo lead, and Google's Gemini rapidly grew its presence. Recently, Samsung Securities analyzed that the latest model, "Gemini 3.0," has reversed the ChatGPT-centric AI landscape as it is being rated the top-performing model. In fact, on LM Arena, a global AI performance benchmarking platform, Gemini 3.0 took the overall No. 1 spot ahead of OpenAI's GPT-5.1, and it reached the top with a 37.5% accuracy rate on the Hard Last Exam (HLE), which measures advanced reasoning ability. At the center of the shift is Demis Hassabis, known as the "father of AlphaGo." A prodigy who ranked No. 2 in the world chess rankings at 13, then a game developer and neuroscientist who went on to win the 2024 Nobel Prize in Chemistry, he is now seen as the "field commander" of Google's AI strategy.
The turning point inside Google came in 2023. Google merged its research organizations DeepMind and Google Brain to launch Google DeepMind and handed full authority to Hassabis. Then, in late 2024, even the Gemini app team was folded under DeepMind, giving Hassabis vertically integrated control spanning research, development and services. The decision dismantled the old structure in which research and product teams operated separately and bound everything under a single chain of command from model development to service launch.
The flagship result empowered under this system is Gemini. Gemini did not evolve with a strategy of merely catching up to ChatGPT; it advanced by fully leveraging Google's scale. The vast unstructured data and user behavior patterns Google accumulated over decades through search, advertising and YouTube became a singular learning asset unique to Gemini. Combined with its native multimodal architecture and cost control on computation using the in-house AI Semiconductor TPU "Ironwood," it boosted both performance and scalability. In particular, Apple's recent selection of Gemini as the generative AI partner for the iPhone, and Alphabet's reclaiming of the world No. 2 market capitalization rank (about $4 trillion), are seen as indicators that Google's technology is becoming the market "standard."
Market response is also shifting quickly. Aggregated global traffic analysis shows that after the second half of 2025, web traffic growth for ChatGPT slowed markedly, while Gemini has shown a relatively steady increase in users. Google has said Gemini's monthly active users (MAU) rose more than 44% in three months, from 450 million in July 2025 to 650 million in October the same year. In particular, observers say it helped that Gemini adopted an "ambient AI" strategy, blending naturally into existing Google services like Search, Gmail and the Android operating system.
These shifts are putting direct pressure on rivals. As Gemini 3.0's pursuit became visible, OpenAI declared a "red code" internally and rushed an early deployment of "GPT-5.2" with enhanced performance last December. CEO Sam Altman has paused AI agent projects such as shopping to concentrate the company's capabilities on defending ChatGPT. However, as infrastructure investments approach $1.4 trillion and push the timeline to break even to 2031 or later, some analysts say financial risks are growing, increasing the burden in a long game against Google, which has its own infrastructure.
Hassabis does not see this contest as a short-term chatbot fight. In several public remarks, he has dismissed claims by OpenAI and others of "Ph.D.-level intelligence" as "baseless," emphasizing that while AI is overhyped in the short term, it remains undervalued in the medium to long term. His view is that there is still a critical gap to reach true artificial general intelligence (AGI). At the same time, he has made clear that AI should expand into a tool that solves hard problems across science and industry, as with "AlphaFold."
This philosophy aligns with Google's business strategy. Google is evolving into a "full-stack AI corporations" that spans cloud, data centers and AI Semiconductor, while combining Generative AI with its existing revenue sources in search and advertising. Unlike competitors that rely on external infrastructure, the ability to control both compute resources and service distribution networks is seen as a decisive differentiator in a long contest.
Ultimately, Gemini's comeback is less the result of mere technical improvement than the accumulation of choices that changed organizational structure and strategic direction. The Generative AI power struggle is now unfolding again around the blueprint drawn by Demis Hassabis.