Gemini 3 set for 2025 launch as Google CEO Pichai manages expectations for frontier model progress

Google is gearing up to launch its next-generation AI model, Gemini 3, with CEO Sundar Pichai confirming a 2025 release during the latest earnings call.

The new model is expected to outpace Gemini 2.5 Pro and close the gap with OpenAI's GPT-5, with a stronger focus on agent-like capabilities for handling complex, multimodal tasks.

Pichai, however, cautioned that Gemini 3 won't deliver overnight breakthroughs. "I'm incredibly impressed by the pace at which the teams are executing and the pace at which we are improving these models," he said, but noted that further advances in frontier models now require more time.

"But it also is true at the same time that each of the prior models you're trying to get better is now getting more and more capable. So I think both the pace is increasing, but sometimes we are taking the time to put out a notably improved model. So I think that may take slightly longer," Pichai said. Historically, Google has released new Gemini versions in December.

Generative AI fuels record growth at Google

Gemini is now central to Google's AI ambitions. Alphabet's quarterly revenue topped $100 billion for the first time in Q3 2025, doubling in five years. "We're firmly in the generative AI era," Pichai told investors.

The Gemini app now counts over 650 million monthly active users, with queries tripling since the last quarter. Still, Gemini trails ChatGPT, which OpenAI reports has 800 million weekly users.

Google Search is seeing major gains from AI features, especially with younger audiences. AI Overviews and AI Mode have driven up search volume, and AI Mode—now available in 40 languages—has hit 75 million daily users. In the U.S., AI Mode queries doubled in the third quarter.

But the new features have sparked controversy. Some website owners say Google's AI paraphrases their content directly in search, cutting into site traffic. Several studies show significant drops in search result clicks.

Google Cloud is a major engine of Alphabet's AI growth. Revenue from "products built on our generative AI models grew more than 200% year-over-year," new cloud customers rose nearly 34 percent, and the order backlog reached $155 billion, up 46 percent from the prior quarter.

More than 70 percent of current customers now use Google AI tools like Gemini, Imagen, and Veo, Google says. Launched in October, Gemini Enterprise already serves over two million subscribers across 700 companies.

Google's custom chips and AI research drive its full-stack approach

Google is betting on a vertically integrated AI stack, combining proprietary chips, in-house models, and custom software platforms to drive higher margins and keep customers loyal over time.

Demand for AI infrastructure is surging. Anthropic is looking to tap up to one million TPUs, and Amin Vahdat, Google's VP and GM of AI and Infrastructure, says every TPU Google has is already fully booked.

"We need more, earlier," Vahdat said, reflecting the urgent demand from customers. The latest Ironwood TPU generation will soon be widely available, along with new cloud offerings powered by NVIDIA's GB300 chips.

On the research side, Google is developing new generative models like the video model Veo 3 and the so-called world model Genie 3. Veo has already been used to generate over 230 million videos, and more than 13 million developers are working with Google's generative AI tools.

Alphabet's self-driving unit Waymo is also scaling up fast. The company aims to launch in London and Tokyo by 2026, with Dallas, Denver, Seattle, and Nashville joining its US network. In San Francisco and San Jose, Waymo has approval for fully autonomous airport rides, and testing is expanding in New York City. New programs like "Waymo for Business" and "Waymo Teens" are designed to reach new user groups.