(Yicai) July 9 -- Jieyue Xingchen Intelligent Technology, the Chinese artificial intelligence unicorn better known as StepFun, is preparing to launch its first agentic AI smartphone, according to sources in the handset supply chain. StepFun also recently raised funds from firms in the consumer electronics chain, as AI competition shifts from cloud-based model training to end-user devices.

The Shanghai-based business, which two years ago became the first Chinese company to release a large language model containing one trillion-parameters, secured investment from Huaqin Technology, one of the world’s biggest original design manufacturers, in its Series B+ funding round earlier this year.

Several businesses in the consumer electronics supply chain, including Longcheer Technology, OmniVision Integrated Circuits Group, and ZTE, are also backing StepFun, according to disclosures about a recent financing round. The startup’s investors now span various segments of the industry, including phone design and manufacturing, mobile imaging technologies, and consumer brands.

Founded in 2023 by former Microsoft executive Jiang Daxin, StepFun specializes in multimodal foundation models, reasoning engines, and AI agents.

“While there’s extensive exploration of LLMs on devices, the smartphone category is still awaiting regulatory licensing,” a source at one Chinese mobile phone maker told Yicai. “From a technical standpoint, however, there are few major obstacles remaining.”

Development of products such as the Doubao AI phone, an experimental agentic device created through a partnership between TikTok-owner ByteDance and hardware manufacturer ZTE, is also gathering pace, the person added.

Several smartphone makers have already identified AI agents as the core direction for next-generation mobile operating systems. Honor has introduced its vision for Agentic OS, while Google has emphasized reconstructing the Android experience around AI agents. As a result, mobile operating systems are emerging as a key battleground in the global AI race.

Unlike overseas AI companies that must build their supply chains from scratch, Chinese AI developers can leverage the country’s mature manufacturing ecosystem to rapidly validate new products and commercialize them through established distribution channels, according to another industry insider.

Rather than relying solely on application programming interfaces and application distribution, LLM developers are aiming to establish more stable user access points through hardware, including smartphones, while generating continuous user data and stronger monetization opportunities.

Industry experts believe that the next round of competition in AI devices will no longer center on simply integrating AI assistants into mobile phones. Instead, it will focus on redesigning operating systems and user interactions with AI agents. As a result, future competition is likely to be driven by agentic AI ecosystems, rather than by new hardware products alone.

Editor: Kim Taylor