<div id="readability-page-1"><div><p>AFTER a blockbuster year for listings, Hong Kong investors were greeted in early January by a rare treat, as artificial intelligence (AI) model developers and chipmakers hit public markets almost simultaneously. With China’s cash-hungry AI companies rushing to tap a buoyant Hong Kong bourse, the city has emerged as a global testing ground for how public markets price the next phase of China’s AI ambitions.</p><p>Within the first two weeks of the year, graphics processing unit (GPU) manufacturers Shanghai Biren Technology and Shanghai Iluvatar CoreX Semiconductor, alongside large-model developers Zhipu AI and MiniMax, raised a combined HK$17.7 billion (S$2.9 billion). The breadth of the offerings – spanning foundational computing hardware and frontier AI models – highlighted investor enthusiasm across the AI value chain.</p><p>International institutions and retail investors piled into the four IPOs, extending the frenzied enthusiasm for Hong Kong listings last year. The demand, however, has sharpened questions over whether the AI boom is inflating valuations for companies that remain deep in the red and, in some cases, lack clear paths to sustainable revenue.</p><p>Still, market observers agree on one point: more AI firms are coming. From large-model developers to chipmakers racing to fill gaps left by US export controls, Chinese mainland companies are lining up to head south – drawn by a flush market and a stock exchange that has moved to smooth their path to listing.</p><h2>A public-market first</h2><p>Zhipu AI and MiniMax have drawn global attention as the first large-model startups anywhere to go public, offering investors an unusually transparent window into a sector that remains largely private elsewhere. Major US model developers such as OpenAI, Anthropic and xAI are still privately held, disclosing only limited operating metrics.</p><p>The two Chinese firms also represent contrasting approaches to commercialising AI models, each of which are now under public market scrutiny.</p><p>Zhipu AI relies primarily on a business-to-business model, offering model-as-a-service products that bundle computing power, development tools and industry-specific applications. According to its prospectus, the company recorded revenue of 191 million yuan (S$35.1 million) in the first half of 2025, but posted a net loss of 2.4 billion yuan, as spending on computing resources and a 657-strong research and development (R&amp;D) team surged.</p><p>MiniMax, by contrast, leans heavily on consumers. More than 70 per cent of its revenue comes from subscriptions, in-app purchases and advertising. As of Sep 30, 2025, its AI-native products had attracted 1.8 million paying users and averaged 27.6 million monthly active users.</p><p>The company’s flagship AI companion apps Talkie and Xingye accounted for more than 20 million monthly active users combined. Talkie, launched overseas in 2023, gained viral attention for celebrity-style chat interactions and briefly ranked among the top five entertainment apps by downloads in the US</p><p>Despite that traction, MiniMax remains unprofitable. In the first three quarters of 2025, it reported revenue of US$53.4 million and a net loss of US$512 million, driven in part by US$180 million in R&amp;D spending.</p><h2>Hardware rally</h2><p>The surge in interest in AI software stocks has been reinforced by a parallel rally in AI hardware, particularly domestic chipmakers positioned to benefit from US export restrictions on Nvidia Corp.’s advanced processors.</p><p>When Nvidia’s chips were effectively cut off from China, domestic players such as Cambricon Technologies helped ignite a re-evaluation of the sector. The chipmaker secured ByteDance as a major client, achieved more than fortyfold revenue growth in 2025, saw its share price jump more than 400 per cent and its market capitalisation top 600 billion yuan, earning it the nickname “King Cambricon.”</p><p>The rally reshaped investor perceptions of domestic GPU-makers. In December, Moore Threads Technology and MetaX Integrated Circuits (Shanghai) both listed on Shanghai’s STAR Market, raising a combined 11.5 billion yuan. Moore Threads quickly rolled out new AI chips and large-scale computing clusters, intensifying competition and accelerating peers’ financing plans. The company now expects its 2025 revenue to more than triple on surging demand for AI computing power.</p><p>China’s Internet companies, long reliant on Nvidia chips for AI training, have begun adopting domestic alternatives. Cambricon and Baidu’s Kunlunxin (Beijing) Technology have emerged at the forefront of the shift. ByteDance has deployed Cambricon chips in advertising recommendation systems and inference workloads for its Doubao large model, one person close to ByteDance told Caixin.</p><p>In 2024, domestic chips were largely backup options, the person said. By 2025, demand surged. While domestic chips still lag Nvidia’s H200 in cost-performance terms, they are already viable for certain use cases, amid what the source described as the “huge gap domestic computing demand.”</p><h2>Why Hong Kong</h2><p>For mainland companies, Hong Kong’s appeal lies in speed and predictability. From listing hearings to IPOs, the entire process for the four AI firms took about three weeks. One executive said the hearing came so quickly that the company did not have enough time to complete all its planned IPO marketing.</p><p>Zhipu AI, MiniMax and Biren Technology all relied on Chapter 18C of HKEX listing rules, which allows “specialist technology” companies that are not yet profitable to go public. Zhipu AI disclosed that it switched from a planned listing on the mainland to Hong Kong in December.</p><p>The number of 18C listings is expected to rise. As of early 2026, 17 of roughly 350 publicly filed IPO applications fell under Chapter 18C, according to PwC Hong Kong, with more such applications expected among the more than 50 confidential filings. A dedicated hotline launched in 2025 allows early regulatory guidance and confidential submissions – features especially valued by technology firms.</p><p>By contrast, mainland IPOs remain tightly controlled, even in policy-favoured sectors. And sharp post-listing price surges in domestic GPU stocks have heightened regulatory caution, reinforcing Hong Kong’s appeal as a more stable alternative, a fund manager at an asset management firm in the city said.</p><p>The attraction extends beyond startups. Baidu has confidentially filed in Hong Kong for an IPO of its AI chip subsidiary Kunlunxin, underscoring the exchange’s growing role in financing China’s AI hardware push.</p><h2>IPO fever and its limits</h2><p>Hong Kong’s IPO market has roared back to life. The “Big Four” accounting firms forecast fundraising could reach up to HK$350 billion this year, up from HK$285 billion in 2025.</p><p>The blockbuster debuts of Biren Technology and MiniMax both attracted more than 400,000 retail investors, while MiniMax’s international tranche was oversubscribed more than 36 times.</p><p>Through Hong Kong’s IPO market, both domestic and international capital is participating in China’s AI investment boom. Li Zhenguo, vice-chairman of UBS Global Investment Banking, noted that a considerable amount of foreign capital had already flowed into the Greater China market in 2025, with further room for allocation.</p><p>Scarcity, however, may complicate valuations, said an investment manager at Fosun Asset Management. Most large-model assets remain embedded within tech giants such as Alibaba Group Holding and Tencent Holdings, leaving few pure-play options. That scarcity partly explains the fervor around Zhipu AI and MiniMax.</p><p>But it also heightens the risk of mispricing. On the eve of Zhipu AI’s listing, an institutional investor said just a roughly 6 per cent grey-market gain in the stock prompted closer scrutiny of IPO valuations. Citic Securities, using OpenAI and Anthropic for comparison, pegged MiniMax’s 2026 price-to-sales ratio at 30x, implying a market value of HK$53 billion – less than half its current capitalisation.</p><p>“There is still no effective valuation framework for large-model companies,” said Chu Yen-Min, chairman of KGI Investment Advisory, warning that rapid technological iteration could quickly render pure large-model players obsolete. Xu Yang, a global partner at Tiger Brokers, added that high valuations, low survival rates and cross-border regulatory risks increase the likelihood of future corrections.</p><p>Wang Fengyu, founder of Oakwise Capital Management, said China’s advantage in AI applications does not guarantee clear winners. Capital, he said, must be patient. “When investing in Chinese AI companies, one must be wary of firms that burn cash endlessly, lack sustainable revenue and have no core technology.” CAIXIN GLOBAL</p></div></div>

02513

06082

00100

<p>Hong Kong has become a key hub for AI companies, with several firms, including Shanghai Biren Technology and Zhipu AI, raising HK$17.7 billion in early January. This surge in IPOs reflects strong investor interest, despite concerns over inflated valuations. Zhipu AI and MiniMax are notable for being the first large-model startups to go public, showcasing different business models. The AI hardware sector is also thriving, with domestic chipmakers benefiting from US export restrictions. Hong Kong's efficient listing process attracts mainland companies seeking rapid access to capital.</p>

In depth: Global capital, Chinese AI converge in Hong Kong