Hong Kong, Jan 14, 2026, 16:33 (HKT)
Shares of Hong Kong-listed Qiniu (02567) jumped more than 7% in intraday trade on Tuesday and were last up 5.97% at HK$0.71, with turnover of HK$1.4 million, after the company disclosed its model-as-a-service, or MaaS, user count had topped 180,000, Zhituo Finance reported. Qiniu has been bundling open- and closed-source AI models including DeepSeek and MiniMax on its MaaS platform, betting it can shift revenue away from basic bandwidth and storage toward higher-value inference API calls, the report said. It pointed to a run of AI catalysts — including talk of a DeepSeek V4 release and a ByteDance AI tie-up around the Lunar New Year gala — and cited Guojin Securities as calling 2026 a turning point for AI applications, with Huajin Securities noting AI is already cutting costs in businesses such as content and marketing. 1
The move matters because investors are trying to pin down who makes money when AI shifts from chips and servers into everyday apps. Early signposts such as user counts and API traffic are getting treated like leading indicators, even when they sit outside formal earnings reports.
For Qiniu, the 180,000 figure is being read as evidence that demand is moving beyond cloud storage into AI services that can be billed per use. That is the kind of transition traders have been waiting to see — and the kind that can fade fast if customers do not pay, or do not stick around.
MaaS is shorthand for “model-as-a-service”, renting access to AI models through a cloud platform, while inference is the moment a model produces an answer to a prompt. In plain terms, companies charge developers when their software calls a model, rather than relying mainly on low-margin data transfer and storage.
In Taiwan, the AI trade has already pushed parts of the market into frothy territory. Market commentator Lin Cheng-yin wrote that many AI-linked stocks have “priced in the next two years’ profits” and warned that when valuations — often expressed through a price-to-earnings ratio — run into the “tens”, any growth miss can trigger a sharp correction; he also noted that only 929 of about 1,900 listed firms were above their 200-day moving average, a common trend gauge, as of Jan. 13. 2
But the same setup that fuels sharp rallies can cut the other way. A delayed product cycle, weaker-than-expected monetisation, or a broader pullback in risk appetite can turn a momentum bid into a quick exit.
In Hong Kong, some big tech names also drew renewed interest early this week, with Hang Seng Tech constituents such as Kingdee and Meituan jumping while Tencent lagged, an EJFQ market note said. The note put Tencent’s price-to-earnings ratio at 23.4, below its average since it became the market’s biggest stock in 2015. It also flagged an overhang from shareholder Naspers’ stated plan to sell shares, even as it said exchange data showed no reductions since August and Tencent has kept buying back stock. 3
The contrast fits the current AI mood: newer names get the spikes, older giants get the valuation debate. For smaller firms like Qiniu, the hurdle is proving that a surge in users becomes durable, paid usage — not just a headline metric.
Traders will be watching the next round of model launches and corporate updates for evidence that the application boom is turning into steady cash flow. The first serious earnings disappointment in the AI complex will be the real test of how much of this is conviction, and how much is heat.
