By Bi Feng
For a generation of Chinese internet users, PPTV was the go-to free service for streaming soccer matches. The platform eventually faded after being acquired by retail giant Suning in 2014 before it could complete an IPO, but its founder never left the table.
Now, Yao Xin is back with a new bet on artificial intelligence infrastructure. His company, PPLabs Technology, also known as PPIO, has just updated its prospectus for a main-board listing in Hong Kong.
The company has emerged as one of China’s largest independent AI cloud computing service providers, positioning itself to benefit from a rapid increase in demand for AI inference — the computing resources required to run large language models and AI applications after they have been trained.
The prospectus shows revenue reached 770 million yuan ($114 million) in the last fiscal year, with a two-year compound annual growth rate of 46.6%. Its AI cloud business expanded more than tenfold over the past year and the company’s platform now handles an average daily token consumption of 1.03 trillion, up from 271 billion at the end of 2025.
Since an angel financing round in 2021 valued the company at $46 million, PPLabs’ valuation has risen to $469 million in its 2025 Series B fundraising. A further $35.2 million Series B-2 preferred share financing was completed in early June.
From video streaming to AI infrastructure
Yao, a computer science prodigy who entered Huazhong University of Science and Technology after winning national programming competitions, developed the predecessor of PPTV, PPLive, in 2004 while still a graduate student.
The service used peer-to-peer technology to improve the streaming of live sports broadcasts at a time when broadband infrastructure was still limited. The more users joined the network, the more efficiently content could be distributed. PPTV eventually attracted 450 million users worldwide and became China’s second-largest video platform, behind Youku.
The technical expertise developed during that period has shaped Yao’s second venture. In 2018, he reunited with former PPTV chief architect Wang Wenyu to establish PPIO, focusing on edge cloud computing.
Rather than building vast data centers and purchasing large numbers of servers, PPIO created a distributed network that aggregates idle third-party computing resources across China. The company describes the model as a way of turning underused computing capacity into a shared infrastructure platform.
Today, PPIO operates more than 4,600 computing nodes across 1,340 counties and cities. Its core asset is not hardware but software that manages and allocates computing resources efficiently.
According to the prospectus, the company’s distributed scheduling system enables graphics processing unit utilisation rates above 75%, compared with industry averages of 40%-50%. It estimates that inference costs can be reduced by more than 80%.
In many ways, Yao’s second entrepreneurial venture mirrors his first. PPTV addressed the challenge of distributing video traffic efficiently; PPIO seeks to solve the problem of distributing AI computing power.
Riding the AI inference boom
The company has benefited from a broader shift in the AI industry. While much attention has focused on training increasingly powerful models, investors are increasingly interested in inference, which generates recurring demand for computing resources every time an AI application is used.
Every chatbot query, image generation request or autonomous AI task consumes inference computing power. As AI adoption spreads, demand for these services is rising rapidly.
PPLabs’ AI cloud computing revenue climbed from just 10.4 million yuan in 2024 to 119 million yuan in 2025. The contribution of AI-related business to total revenue increased from less than 2% to 15.5%. The growth in token consumption has been even more striking. Average daily token usage surged from 271 billion in December 2025 to 1.03 trillion by April 2026.
The company says it had more than 574,000 registered developers on its platform by April, suggesting that a growing ecosystem of AI application developers is relying on its infrastructure.
Betting on the age of AI agents
PPLabs is now attempting to position itself beyond cloud computing and as a provider of infrastructure for AI agents. Industry executives increasingly expect agents to generate substantially higher computing demand than conventional AI chatbots. A single agent may need to search for information, access tools, analyse data and generate content, triggering dozens or even hundreds of inference requests in a single workflow.
PPLabs argues that future AI infrastructure will require not only computing power but also systems for orchestration, memory management, sandbox environments and task scheduling.
The opportunity is substantial, but so are the risks. China’s AI infrastructure market is becoming increasingly competitive, with major cloud providers including Alibaba Cloud, Tencent Cloud and Baidu AI Cloud investing heavily in AI services. The industry is also facing mounting price competition.
PPLabs remains loss-making, while margins are under pressure and the company continues to depend heavily on a small number of major customers. Significant investment will be required to expand its computing network and maintain technological competitiveness.
For Yao, however, the IPO represents another attempt to capitalise on a technological shift. Having ridden the rise of online video two decades ago, he is now betting that the next great infrastructure race will be driven not by streaming media but by AI. Whether that wager proves successful may depend on how central computing power becomes in the emerging AI economy.
Source:
LatePost
