Cover | Produced by Pencil News

An AI startup, generating less than 18 million yuan in half-year revenue and suffering losses exceeding 150 million yuan, is on the verge of securing a 300 million yuan investment from an A-share listed company.

On September 29, Wangsu Science & Technology unveiled its plan to invest 300 million yuan in video generation model firm Sand.ai, acquiring a 4.3963% stake post-transaction. This investment values Sand.ai at approximately 6.8 billion yuan.

Why would Wangsu Science & Technology, a 26-year-old internet infrastructure provider, pour 300 million yuan into a video model company that is still experiencing significant cash burn?

The answer may lie in Wangsu's ongoing business transformation.

- 01 - A Young AI Company Valued at 6.8 Billion Yuan

Sand.ai's founder, Cao Yue, is a prominent technical talent in China's AI startup landscape.

Cao holds a PhD from Tsinghua University's School of Software and previously worked as a researcher at Microsoft Research Asia. His contribution to the Swin Transformer research earned the 2021 International Conference on Computer Vision (ICCV) Best Paper Award—the prestigious Marr Prize.

In February 2023, Meituan co-founder Wang Huiwen announced his venture into large model entrepreneurship, aiming to create "China's OpenAI."

According to LatePost, Wang Huiwen met with leading AI researchers, seeking recommendations for the next talent to engage. A month later, he discovered Cao Yue, then 31 years old.

Cao subsequently joined Lighthouse AI as the co-founder for algorithms. In 2023, Wang Huiwen stepped down from company management due to health issues, and Lighthouse AI was acquired by Meituan.

In January 2024, Cao Yue embarked on his second entrepreneurial journey with Sand.ai. This time, rather than directly competing with OpenAI in the general language model market, he focused on video generation.

Sand.ai's business model comprises two key components: developing large video generation models that enable AI to create dynamic content from text or images, and commercializing these models through applications that charge creators.

In January 2024, Sand.ai launched VidMuse, an AI tool for music video creators that generates visuals driven by music. The company revealed on its website that within two months of launch, VidMuse's annualized recurring revenue (ARR) surpassed $10 million.

In August 2024, Sand.ai introduced and open-sourced the MAGI-2 Preview video model. Featuring 114 billion parameters but utilizing a Mixture of Experts (MoE) architecture, it activates only about 6 billion parameters per token processed.

However, its actual revenue remains modest.

According to Wangsu's announcement, Sand.ai generated 17.398 million yuan in revenue in the first half of the year, with a corresponding loss of 152 million yuan. For a company valued at approximately 6.8 billion yuan, current revenue clearly falls short of supporting traditional financial valuation methods.

Investors are betting on its potential to become a key player in the AI video market.

Sand.ai has previously attracted investments from institutions such as Capital Today, Matrix Partners China, Sinovation Ventures, and Source Code Capital. Wang Huiwen and Kuaishou founder Su Hua are also among its investors.

This time, Wangsu brings not only capital but also a global content (potential) distribution network to the table.

- 02 - The Industry Leader Brought Low by Price Wars

Wangsu Science & Technology and Sand.ai represent two distinct eras of China's internet industry.

The former was established in 2000, on the eve of the dot-com bubble burst; the latter emerged in 2024 amid the global surge in large model startups.

Wangsu's founding story is legendary.

Co-founder Chen Baozhen, born in 1943, worked at the Electronic Instrument Factory of Fujian's Jimei University for about 30 years. In 2000, after retiring, she and Zhou Aijun each invested 1 million yuan to establish Wangsu's predecessor, initially focusing on internet data center services.

In 2001, Liu Chengyan, former COO of China Wanwang, joined the team. The company gradually shifted its focus to CDN (content delivery) and went public on the ChiNext board in 2009.

The growth of internet video sites, gaming, and e-commerce fueled rapid expansion. By February 2014, Wangsu's stock price surpassed Kweichow Moutai, making it one of the highest-priced stocks on the A-share market. Chen Baozhen became a well-known figure in China's early internet entrepreneurship wave.

The turning point came in 2015.

In May 2015, Tencent Cloud CDN announced a 25% price cut, followed by Alibaba Cloud, triggering a price war. 2017 was particularly brutal: in March, Alibaba Cloud CDN reduced prices by 35%; in November, Alibaba Cloud cut overall prices by 25%; ten days later, Tencent Cloud CDN slashed prices by up to 47%. Cloud computing companies supplanted CDN providers "like trains replacing horse-drawn carriages."

Wangsu lacked the confidence to engage in a price war. Alibaba and Tencent did not rely on CDN for profit but used it to acquire customers, offsetting losses with revenues from public cloud, advertising, and e-commerce. Wangsu's single-focus business model could not compete.

Wangsu's CDN gross margin plummeted from around 47% to 35% by mid-2017 and now stands below 16%. Net profit was expected to drop 32-40% year-on-year in 2017. Between 2017 and 2018, Chen Baozhen and Liu Chengyan, as parties acting in concert, cashed out approximately 1.488 billion yuan. In January 2019, Wangsu sold its IDC business (Xiamen Qinhuai).

A former CDN king with a market cap of 53 billion yuan and a stock price of 141 yuan was thus rewritten by cloud computing and price wars.

- 03 - The Veteran Seeks a Comeback

Understanding this history reveals the true significance of the 300 million yuan investment.

First, consider Wangsu's current position. In the first half of 2026, revenue reached 2.318 billion yuan (+13.11% YoY on a like-for-like basis), with net profit attributable to shareholders of 348 million yuan. Total assets stood at 11.777 billion yuan, with operating cash flow of 365 million yuan in the first half. Three billion yuan represents nearly half a year's operating cash flow—not a small sum, but manageable.

The core business continues to struggle: in H1 2026, CDN and edge computing revenue was 1.368 billion yuan, down 9.52% YoY, with a gross margin of just 15.85%. In 2025, this business generated 2.895 billion yuan in revenue, down 15.58% YoY. Growth now comes from security and value-added services, with a ~74% gross margin, 790 million yuan in revenue, and 22.1% YoY growth—but at one-third of total revenue, it cannot shoulder the load alone.

This investment signals that with its core business capped, Wangsu must find new uses for its 2,800+ edge nodes.

Wangsu's strongest asset is its 2,800+ edge computing nodes deployed across 70+ countries. Previously used mainly for content delivery, these nodes are being upgraded to store, compute, and perform AI inference. In H1 2026, it launched the "Edge AI Execution Engine" and edge AI gateways, and in July, partnered to build a city-level token factory in Changzhou.

The question remains: who will utilize these nodes for high-value AI inference? That's where Sand.ai comes in. Video generation is notoriously resource-intensive—exactly what Wangsu's 2,800 nodes offer.

However, Wangsu's investment does not guarantee that Sand.ai's video models will immediately deploy on Wangsu's global servers.

Wangsu's announcement was cautious, stating that the two sides will explore business cooperation in audio-video inference, low-latency edge inference, and other areas.

As of the September 29 announcement, no formal investment agreement had been signed, and no joint commercial orders were disclosed.

For this veteran tech company, the 300 million yuan investment in Sand.ai—acquiring ~4.4% equity—is not merely a financial bet but an attempt to enter the AI model industry chain.

One figure warrants long-term observation: in H1 2026, Wangsu's CDN and edge computing gross margin was just 15.85%, while security services reached 74.03%.

If AI can deliver similar high-margin revenue, this investment may prove more valuable than its equity stake suggests.

This article does not constitute investment advice.