PPIO eyes next AI frontier with plans to boost computing power

The provider of cloud services wants to upgrade its systems as AI moves into the mainstream, demanding greater processing capacity and speed

Key Takeaways:

• The loss-making company has filed for an IPO to raise fresh funds to meet the anticipated growth in AI-related demand
• It specializes in localized “edge” processing for faster response times, with revenues jumping 56% last year   

By Lee Shih Ta

As AI technology takes off, more computing power will be needed to apply generative learning to practical tasks and new contexts. This process, known in the industry as inference, could drive exponential demand for cloud services.

Gearing up for a fight over processing bandwidth, a Chinese provider of cloud services has applied to list on the Hong Kong Stock Exchange, with global ambitions to help empower the next phase of the AI revolution.

In its prospectus, PPLabs Technology Ltd., the parent of cloud services provider PPIO, cited plans to upgrade its AI-related services and develop its overseas presence in distributed cloud computing. The company describes itself as as pioneer of cloud services in China for AI tasks, ranking second among independent providers for average daily use.

The daily consumption rate in May reached 141.9 billion tokens, the system used by the computing industry to measure and price its operations.

Second chances

Company founder Yao Xin is no stranger to the tech industry. As a graduate student at Huazhong University of Science and Technology, Yao founded China’s earliest online video platform in 2005. PPLive Media, later rebranded as PPTV, made its debut a year before YouTube, raising $700 million in capital before being sold to retailer Suning in 2014. At that point Yao paused his entrepreneurial activities.

Returning with the launch of PPIO in 2018, he focused the new business on an aspect of cloud computing that was not yet dominated by tech giants. The company set out to specialize in distributed edge computing, which aims to process data close to where it is generated.

The moves reflect shifting dynamics in cloud computing with the transformative advent of artificial intelligence. The costs of rolling out extrapolative AI systems and the challenge of allocating computing resources to them are causing bottlenecks. A capacity battle is brewing over the responsiveness to AI requests, in which edge computing can offer a proximity advantage.

Take the example of a Beijing user interacting with an AI assistant. If the task is handled at a U.S. data center, response delays are hard to avoid, even with optimized network transmission. However, a request routed via a node in Beijing or Hebei could be processed more like a real-time experience, which is particularly critical for voice, video, or gaming applications.

According to China Insights Consultancy, the edge cloud computing market in China is set to reach 37 billion yuan ($5.16 billion) in revenue by 2029, with a compound annual growth rate (CAGR) of 22.9% from 2024 to 2029. The global market was projected to achieve a CAGR of 68.5% over the same period, reaching 427.7 billion yuan in 2029.

Companies offering edge cloud services could become vital links in the AI industry chain, connecting infrastructure providers such as telecom operators with suppliers of processing hardware while packaging and distributing computing power along the way, and supporting AI developers.

PPIO has built a distributed computing network covering more than 1,200 cities and 4,000 nodes worldwide through its own infrastructure and partnerships.  Using a scheduling platform developed in house, the network enables AI tasks to be automatically routed to the nearest available resource.

An Airbnb model?

The business monetizes the market in three ways. Firstly, computing power resources are provided to developers on demand, with billing based on duration or the tokens consumed. This income stream accounts for around 70% of revenue. In edge node services, idle hardware from enterprises or developers is co-opted into the node network, generating revenue through shared computing power. The company also offers customized AI applications, including services for small-model reasoning and systems to distribute content to nearby networks.

With its business still in the expansionary phase, PPIO has yet to break even. Still, it compares favorably to some cloud service startups, with relatively contained losses and a gross profit margin above 10%.

Figures in the IPO filing show that its turnover and losses have both been rising. The company logged revenues of 286 million yuan in 2022, increasing to 358 million yuan a year later and jumping nearly 56% in 2024 to 558 million yuan, a CAGR over the period of nearly 40%. Gross profit margins were 16.1%, 17.7% and 12.3%, while losses amounted to 85 million yuan, 189.4 million yuan and 293.5 million yuan, totaling around 568 million yuan for the three-year period.

The way PPIO aggregates idle processing resources and uses multi-node scheduling for AI tasks is similar to Airbnb approach of utilizing idle rooms. The PPIO model is suited for use with AI agents and for visual tasks or low-latency interactive scenarios.

However, distributed architectures also face challenges when it comes to regulation, standardization and service stability. Market competition is also intensifying, as major tech players step into the arena. Aside from direct rivals such as UCloud, Alibaba (BABA.US; 9988.HK), Baidu (BIDU.US; 9888.HK)  and Kingsoft Cloud (3896.HK) are building up their own AI computing power and edge nodes. To maintain its position, PPIO would need to stay ahead of the game for platform experience, pricing and international reach.

To that end, PPIO does need external financing to keep expanding and to avoid a liquidity crunch, having around only 114 million yuan in cash on hand and a net debt of 696 million yuan.

The streaming service PPTV, which never went public, could be viewed as a missed opportunity for founder Yao. The cloud computing firm PPIO offers a second chance at stock market success, if the company can effectively plug into the surging demand for AI computing.

To subscribe to Bamboo Works weekly free newsletter, click here