Baidu announced a smart speaker and a cute-looking interactive home robot at an event in Beijing today. 

Jesse Lyu, general manager of Baidu’s Intelligent Hardware Unit, preceded his product launch presentation at the Baidu World conference with a tribute to the iPhone. He noted that every smartphone that went on the market after it looked pretty much the same, and that’s because the iPhone was the “definitive product.”

Minutes later, Lyu, who is also the founder of Raven Tech, a startup acquired by Baidu in February, revealed his vision for a definitive product in the age of AI: Raven H, a smart speaker that looks like a neat stack of square orange, red, blue, and green panels. Lyu also revealed the Raven R, a similar device that can also move around on six axes.

The devices are Baidu’s first AI-focused hardware products. And both use DuerOS, Baidu’s conversational AI system. The Raven H will cost 1,699 yuan ($257). A price hasn’t been announced for the Raven R. Each device can be used to look up information, play music, or order a taxi. But the Raven R will move in response to a user and dance along to music.

The company joins a growing number of Chinese tech companies making forays into hardware with built-in AI. Alibaba released a smart speaker called Tmall Genie in July, and Tencent has announced a similar product called Xiaowei. JD.com, a major Chinese e-tailer, also released a smart speaker, called DingDong, in 2015. And Megvii, a Chinese company that runs the largest face-recognition technology platform in the world, called Face++, offers a face-recognition module that can be connected to computers via USB. Chinese companies that manufacture traditional home appliances such as TVs and fridges are also looking to make their products smarter by adding voice control features.

All these new products need processing chips powerful enough for the computing required by AI algorithms, yet small enough to be embedded in mobile and wearable devices.

The increasing demand for AI hardware could fuel a makeover of China’s chip industry. “The age of AI is a new beginning,” said Shuran Wei, CEO of RDA Microelectronics, a Shanghai-based company that produces smartphone chips, during a panel discussion after the launch of Raven H.

Indeed, both company executives and researchers in China are hopeful that the country’s chip industry could ascend to a leading position in the new wave of chip development focused on deep learning. Currently China spends over $200 billion a year on importing integrated-circuit chips—a statistic often cited by Chinese leaders as both a reality check and an incentive for Chinese researchers and companies to work harder. In 2014, a group of government agencies and investors led by the Ministry of Industry and Information Technology established a fund for integrated-circuit development.

There is already some progress. Cambricon, a Beijing-based AI chip company, released a series of AI processors that can be used in security cameras, drones, smart speakers, and cloud computing in early November. The company aims to have its smart processors embedded in over a billion devices after three years. A research team called Reconfigurable and Programmable Intelligent Device at Tsinghua University has also designed a general-purpose neural-network processor that can perform different kinds of AI tasks, such as voice and face recognition.

Companies that don’t have the expertise to manufacture AI processors yet are navigating this new opportunity in a different way. RDA Microelectronics, a Shanghai-based company, currently makes a chip that can record a person’s voice and then send that information to a cloud platform that stores audio files. It’s now widely used in a gadget that can tell stories to children through voice commands.

Hardware should be an important element in an AI company’s business model, says Shouyi Yin, the vice director of Tsinghua University’s Institute of Microelectronics, who also leads the team that developed a neural-network processor there.

But there are challenges in developing chips for use in these hardware products. A big one is striking a balance between energy efficiency and computing power, since many of these chips will have to be embedded in tiny wearable devices. Another is devising chips that let AI do completely new things. “There are many different AI techniques,” says Yin. “We need new frameworks and new chip designs.”