September 26, 2026

Google CEO Sundar Pichai : Next-Generation AI Model Gemini 3 to Launch This Year

November 3, 2025
ModeZone

November 2 — According to The Decoder, Google is preparing to release its next-generation artificial intelligence model, Gemini 3, and CEO Sundar Pichai confirmed during the company’s latest earnings call that the model will be launched in 2025.

Gemini 3 is expected to surpass the current Gemini 2.5 Pro, further narrowing the performance gap with OpenAI’s GPT-5, while placing a stronger emphasis on “agent-like” intelligent capabilities to better handle complex, multimodal tasks.

However, Pichai cautioned that Gemini 3 will not deliver immediate, groundbreaking leaps. “I’m very encouraged by the team’s execution speed and the pace of model iteration,” he said. “At the same time, we recognize that advancing frontier models now requires more time.”

He added: “Each generation of models has continuously improved in capability. While the overall pace of progress is accelerating, it sometimes takes longer to release major upgrades. This means certain updates may arrive later than expected.” Historically, Google has often introduced new Gemini versions in December.

Gemini has become the cornerstone of Google’s AI strategy. In Q3 2025, parent company Alphabet’s quarterly revenue surpassed $100 billion for the first time, with Pichai telling investors: “We are clearly in the era of generative AI.”

Currently, Gemini’s apps boast over 650 million monthly active users, with query volume doubling quarter-over-quarter. Despite this, Gemini still trails behind ChatGPT, which OpenAI reports now has 800 million weekly active users.

Google’s Search business has also seen significant growth driven by AI features, particularly among younger users. AI Overviews and AI Mode have boosted search engagement, with AI Mode now supporting 40 languages and reaching 75 million daily active users. In the U.S. market, AI Mode queries doubled year-over-year in Q3.

However, these new features have sparked controversy. Some website operators claim that Google’s AI directly summarizes or rewrites original content in search results, causing sharp declines in site traffic. Studies show that click-through rates on traditional search links have fallen significantly.

Meanwhile, Google Cloud has become a major growth driver for Alphabet’s AI business. The company reported that revenue from AI-based products grew more than 200% year-over-year, with new cloud customers increasing by nearly 34% and a backlog reaching $155 billion, up 46% from the previous quarter.

Google noted that over 70% of existing customers now use its AI tools, including Gemini, the Imagen image generation model, and the Veo video generation model. The Gemini Enterprise edition, launched in October, is already used by over 700 enterprises worldwide, with more than 2 million subscribers.

The company continues to build an end-to-end, vertically integrated AI technology stack, combining in-house chips, internally developed large models, and customized software platforms to improve margins and increase customer retention.

Demand for AI infrastructure is surging. Reports indicate that AI startup Anthropic plans to utilize up to 1 million Google TPUs (Tensor Processing Units). According to Amin Vahdat, Google’s VP and GM of AI and Infrastructure, all available TPUs are now fully booked.

“We need to deliver more resources faster,” Vahdat acknowledged, reflecting the intense demand for compute power. The next-generation Ironwood TPUs are set for large-scale deployment, alongside a new cloud service line powered by NVIDIA GB300 chips.

On the R&D front, Google is advancing several new generative models, including the Veo 3 video generation model and Genie 3, a so-called “world model.” So far, Veo has generated over 230 million videos, and more than 13 million developers worldwide are using Google’s generative AI tools.

Additionally, Waymo, Alphabet’s autonomous driving division, is accelerating its global expansion. The company plans to enter the London and Tokyo markets by 2026 and add Dallas, Denver, Seattle, and Nashville to its U.S. network. Waymo has already obtained licenses for fully driverless airport transfers in San Francisco and San Jose, and testing in New York City continues to expand. To reach new customer groups, Waymo has also launched “Waymo for Business” and “Waymo Teens” programs.

The above content is compiled by ModeZone, a fashion and entertainment magazine.

You May Also Like