Google reports 14x growth for Gemini API, plans next-gen assistants for 2025

Google’s Gemini AI services are experiencing substantial growth, with API usage soaring 14 times over the past six months, CEO Sundar Pichai revealed during the company’s quarterly earnings call.

Pichai revealed that Google has integrated Gemini into its core products, including Maps and Search, potentially reaching more than two billion users. Search alone reaches over a billion people. The standalone Gemini app and its integration with Android Assistant have also seen strong user growth, Pichai said.

What these numbers tell is that Google has a clear lead in AI distribution. What they don’t say is how happy people are with these features, which would be the basis for future monetization.

Google’s cloud division is already reaping the rewards of the AI boom, with revenue jumping 35 percent to $11.4 billion. That includes both AI and third-party services, and the division’s operating income jumped from $270 million to $1.95 billion year over year.

Google Services, which includes search, reported a solid 13 percent increase in revenue to $76.5 billion. Despite the hype around AI-powered search competitors like Perplexity and OpenAI’s SearchGPT (which is still in beta), Google hasn’t seen any significant impact on its search dominance yet.

Google teases Astra-inspired AI experiences for 2025

Pichai teased new AI experiences for 2025, drawing inspiration from “Project Astra.” Google unveiled Astra last summer as its “vision for the future of AI assistants” – a multimodal AI capable of processing text, video, and audio in real-time while seamlessly understanding and interacting with the real world around it, not just confined to a browser.

According to media reports, Google is gearing up to release Gemini 2.0 and a potential app and web navigation agent codenamed “Jarvis” in December. While Gemini 2.0 reportedly won’t bring major performance improvements, the Jarvis agent could pave the way for more advanced Astra-like assistants next year. Jarvis seems more akin to Anthropic’s AI agents, which can unreliably do very basic computer stuff in interfaces designed for humans.

Pichai also revealed that AI generates more than 25 percent of Google’s new code, which is then reviewed and approved by human engineers. Amazon has reported similar efficiency gains for routine coding tasks, claiming that AI-driven coding has saved thousands of development hours.

Take these claims with a grain of salt until the companies actually provide concrete data and code quality assessments. Both companies have a vested interest in promoting their AI models and cloud services to capitalize on the growing AI trend, so some skepticism is warranted.