In 2026, Hong Kong is entering an IPO cycle led by artificial intelligence companies.
Compared with last year’s rush of hard tech companies and A-share dual listings, the main theme for Hong Kong IPOs this year has clearly shifted toward AI. The market is also showing greater tolerance for high valuations and stronger investor enthusiasm.
Since the start of the year, the Hong Kong Stock Exchange has seen a succession of AI companies reach RMB 100 billion (USD 14.7 billion) in market capitalization. Large model developers such as Zhipu AI (also known as Z.ai) and MiniMax have risen to valuations in the hundreds of billions of RMB. Xunce Technology once surged 500% this year to join the RMB 100 billion club, while chip companies such as Iluvatar CoreX and Biren Technology have also crossed that threshold.
From a financial perspective, these companies’ revenue and profit are still limited. The more important change is that AI commercialization is moving from expectation to delivery. Compute, data, and models are beginning to connect more deeply with real-world demand, and the market is pricing in future performance earlier.
These IPOs are setting a precedent for leading industrial AI companies to enter the capital markets. The filings of companies such as SmartMore have not only expanded Hong Kong’s AI landscape, but also brought a more practical question to the foreground:
- How much auditable and scalable value can AI create in frontline industrial scenarios?
- And when large model and chip companies already enjoy massive valuations, how much can this IPO wave actually deliver?
Is industrial AI ready for scaled deployment?
Introducing AI into production processes is a key issue in the intelligent transformation of manufacturing. But this is not simply a matter of experimenting with technology. It is also a response to demographic changes and rising labor costs.
For enterprises, replacing labor with machines and using algorithms to improve production yield are moving from nice-to-have upgrades to necessities. A Cisco research report said the use of AI in industrial operations has moved beyond the experimental stage and into broad adoption, with 59% of manufacturers already beginning deployment at scale.
However, the word “deployment” obscures important differences. For most manufacturers, deployment remains concentrated in standardized scenarios such as localized inspection and predictive maintenance. Only a minority have moved into core links such as process control and production scheduling optimization. This is also why revenue for Chinese industrial AI companies is generally stuck around the RMB 100 million (USD 14.7 million) threshold, while only a small number of leaders can reach RMB 1 billion (USD 146.8 million).
The problem is not that the market is small. Rather, industrial production lines demand near-absolute certainty, while the probabilistic output of large models naturally conflicts with that requirement. Combined with fragmented production line data and companies’ concerns over network and data security, pushing AI into the harder phase of implementation has proved more difficult than expected.
As a result, AI breakthroughs have first concentrated in high-end manufacturing fields with stronger infrastructure and higher levels of automation. A survey referenced by 36Kr confirms this distribution: AI adoption is leading in sectors such as semiconductors and electronics, automotive, and energy and power. These are areas that already have stronger data foundations and engineering capabilities. They also have greater motivation to use AI to solve complex production and R&D problems.
As scaled deployment begins in these sectors, the business model for industrial AI is also changing. It is moving from the early validation stage, dominated by customized projects, toward replicable delivery at scale.
This scale still depends on leading customers and specific scenarios, and remains some distance from broad adoption. But that also means the market still has significant room to grow.
This change is gradually showing up in company operations. The industry is moving from point validation to commercial expansion, and revenue growth among leading companies has become the clearest signal.
Take SmartMore as an example. Its revenue reached RMB 1.086 billion (USD 159.4 million) in 2025, while the number of customers it served exceeded 730. Revenue contribution from its top five customers fell from 39.3% to 22%, indicating a more diversified customer structure.
By industry, SmartMore’s scaled implementation is concentrated in high-end manufacturing, including consumer electronics, new energy, precision manufacturing, and rail transit. Its customers include leaders in their fields, including Tesla, Luxshare Precision, GoerTek, BOE Technology, and Kedali.
These companies operate in industries that generally feature complex manufacturing processes, relatively high automation levels, and strong sensitivity to production yield and efficiency. That may explain why they have become the first scenarios in which industrial AI has delivered value.
Are industrial AI agents the practical answer?
In an industrial value chain, production and manufacturing are the links where AI can most readily deliver value. Efficiency gains and cost reductions can often translate directly into corporate profit.
The problem is that most current industrial AI deployments remain at the inspection level: AI detects a defect, and a human handles it. This makes it difficult for AI to enter the core links of production lines. To truly reshape yield and efficiency curves, AI must be able to directly call equipment, adjust parameters, and execute actions.
This is the question industrial AI agents are trying to answer.
Put simply, an industrial AI agent is a system that can complete perception, reasoning, and execution in industrial scenarios. With large models at the core, it combines industry data and business processes, allowing it to understand production problems and call systems or equipment to complete tasks. Its form can be a large model, an industrial software system, a robot, or an integrated software-hardware solution.
Consider a production line focused on mounting smartphone motherboards. Traditional AI can tell you that the weak solder joint rate in a batch is 3% higher than the standard, after which engineers stop the line and debug the issue. An industrial AI agent can compare the situation with historically optimal parameters in real time, automatically adjust solder paste thickness and reflow soldering temperature for the next board, and, after improvements across several consecutive boards, lock the new parameters into the production line control system.
The entire process requires no human intervention. This is the essential difference between analysis and execution, and it is also why factories are willing to keep paying.
Among international vendors, industrial AI agents are becoming a consensus, though the paths differ:
- Siemens, for example, has launched the engineering agent Eigen, which can directly participate in engineering configuration and control logic generation. It is more of a software-based engineering agent.
- ABB starts from hardware robots, embedding AI capabilities into the execution end so they can complete operations in real environments.
- Cognex, meanwhile, integrates AI capabilities into machine vision equipment, making devices more intelligent by improving perception and local decision-making.
Despite the different paths, the common thread is clear: all of them are pushing AI from an analytical tool toward an execution unit in the production process.
Domestic vendors in China are more inclined to advance through an integrated approach. Using SmartMore as the example, it has built a system spanning AI infrastructure at the foundational layer, large models and a general-purpose industrial platform at the middle layer, and edge perception and robotics terminals at the application layer. Robotic products equipped with the industrial multimodal large model IndustryGPT are gradually becoming its growth core.
Its financial data reflect the same shift. Between 2023–2025, SmartMore’s revenue increased from RMB 485 million (USD 71.2 million) to RMB 1.086 billion, representing a compound growth rate of about 50%. Over the same period, revenue contribution from its industrial AI agents, including robots, edge AI sensors, and agentic software systems, rose from 62.4% to 78.5%, showing continued concentration toward agent-related businesses.
A closer look shows that revenue contribution from the robotics business rose from 29% to 40.1%, reaching RMB 436 million (USD 64 million) in 2025. Its three-year compound growth rate exceeded 70%, suggesting that the commercial center of gravity for industrial AI is shifting toward execution-oriented agents.
Another signal worth noting is the increase in value per customer. SmartMore’s single-customer revenue from robot products rose from RMB 1.826 million (USD 268,056.4) in 2023 to RMB 3.63 million (USD 532,883.1) in 2025. This reflects rising customer acceptance of related products. It may also suggest that industrial AI deployment is becoming deeper, rather than simply expanding through customer count.
For industrial AI participants, once business direction becomes clearer and revenue begins to grow, the more practical question is whether cash flow can support long-term investment and continued operations.
Industrial AI is a long-cycle, high-investment sector. It requires sustained software and hardware R&D spending, as well as long-term accumulation of process and scenario know-how in vertical industries.
According to China Insights Consultancy (CIC), participants in China’s industrial AI agent market mainly fall into two categories: traditional overseas industrial giants and local startups. The former rely on established businesses for stable cash flow, while the latter depend more on financing and gradual commercialization.
As a result, cash flow quality has naturally become one of the key indicators capital markets use to evaluate industrial AI companies in China.
With a 5.8% share of China’s industrial AI agent market, SmartMore offers capital markets a case study for whether companies in the sector can generate internal cash flow.
According to its prospectus, SmartMore appears on the surface to face pressure from widening losses. Its accounting net loss expanded from RMB 546 million (USD 80.2 million) in 2023 to RMB 991 million (USD 145.5 million) in 2025. But tracing the causes of those losses shows that they are largely due to non-cash items under accounting standards.
There are two main reasons for the wider accounting loss.
The first is changes in the fair value of preferred shares, which amounted to negative RMB 239 million (USD 35.1 million) in 2025. Although this appears as a loss, it is actually the result of the company’s valuation rising, which made the preferred shares held by earlier investors more valuable. Accounting standards require this to be recorded as an expense. The entry does not affect actual cash flow, but it can alarm readers who focus only on the income statement.
The second is share-based payment expenses. In 2025, the company recorded RMB 475 million (USD 69.7 million) in equity incentive expenses, mainly to retain and motivate its core management and technical teams.
Such financial effects are common among technology companies during the listing process. According to 36Kr, several AI companies previously recorded similar large losses because of valuation increases. Excluding these non-cash factors, SmartMore’s adjusted net loss, which better reflects its actual business cash generation capability, narrowed from RMB 390 million (USD 57.3 million) in 2023 to RMB 270 million (USD 39.6 million) in 2025.
More noteworthy than narrowing losses is the operating leverage released as sales scale grows.
The industrial AI agents SmartMore delivers to customers combine software and hardware. The company must procure a wide range of hardware components, vision and optical components, chips, and other raw materials from upstream suppliers. In the early stages of development, small procurement volumes meant it lacked bargaining power.
But as sales scale rose, SmartMore established stable cooperation with upstream suppliers and brought hardware costs under control. At the same time, software R&D has very low marginal costs, and increasing sales scale can effectively dilute R&D expenses. The prospectus shows that between 2023–2025, SmartMore’s gross margin rose steadily from 30.5% to 37.3%.
Scale effects also flowed through to the expense side. As the revenue base expanded, relatively fixed administrative and R&D expenses were diluted. Its adjusted operating expense ratio fell sharply from 113.6% to 64%. The improvement in front-end gross margin and the decline in back-end expense ratio together form the underlying logic of the company’s continued loss narrowing.
Overall, rising revenue scale and shrinking losses have become the main financial theme for leading industrial AI companies in China.
Will capital markets recognize the value of industrial AI?
The primary market is accelerating its bets on industrial AI. UBS data indicate that AI and machine learning deals accounted for 38% of global venture capital invested in the industrial sector in the first half of 2025, even as overall industrial-linked VC investment remained relatively flat since 2022. Within those AI and machine learning deals, investment rose 268% year-on-year during the same period.
The more important question is whether that enthusiasm can be priced in the secondary market. Hong Kong stocks have historically placed greater weight on fundamentals and shown limited tolerance for concept speculation. In the current global AI rally, however, fear of missing out on AI gains is temporarily changing that constraint. It has also created a relatively loose listing window for industrial AI companies with the dual narrative of large models and robotics.
But short-term enthusiasm does not provide sustained valuation support on its own. Based on current data, industrial AI has already begun to move beyond the purely conceptual stage. CIC expects the global industrial AI agent market to grow at a compound rate of 35% between 2025–2030. Leading companies represented by SmartMore have also produced historical compound revenue growth of more than 50%, alongside narrowing losses.
Filing for an IPO is only the starting point. For industrial AI companies that have just begun their Hong Kong listing journeys, the market’s real focus is not historical performance. It is whether growth and loss-narrowing trends can continue to be verified across a series of milestones: filing, listing, and the first quarterly report.
Specifically, the market will examine three progressively deeper questions:
- First, can growth continue to move from leading customers into a broader base of midsized manufacturers?
- Second, are narrowing losses supported by real internal cash generation?
- Third, can scale effects translate into hard financial metrics and continue to be delivered after listing?
KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Xiao Xi for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.81 = USD 1 based on estimates as of May 21, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.