Global industrial transformation is gathering pace, and the rules of the innovation competition are changing.
The 2026 Summer Davos in Dalian, Liaoning province, themed “Innovating at Scale”, will focus on improving institutions, guiding capital and shaping forward-looking policies. Its core objective is to turn technological potential into economic growth, business development and job creation.
As artificial intelligence penetrates industries and society, a technological breakthrough alone is not enough to secure a lasting competitive edge.
The true measure of innovation lies in whether the technology can be embedded in real-world scenarios, replicated across industries and transformed into a productive force.
For years, innovation has been judged at the front end of the value chain: by algorithm upgrades, model improvements, hardware breakthroughs, academic papers, patents and technical parameters.
These indicators matter because basic research and original innovation are the wellspring of the entire innovation system. But technological progress defines only the ceiling of innovation, not its economic value.
A breakthrough in the laboratory does not always translate into growth unless it reaches the market.
In recent years, global investment in AI research has soared and new advances have emerged. Yet many frontier achievements remain confined to laboratories rather than empowering the real economy.
Technology creates value only when it finds practical use and can be deployed at scale. This is the key difference between point innovation and scaled-up innovation. Point innovation delivers a breakthrough from zero to one, focusing on technical verification and parameter optimization.
Scaled-up innovation completes the journey from one to many, adapting mature technologies to industrial needs, reshaping production models and creating economic value at a broader scale.
AI, in particular, depends on industrial scenarios. General-purpose models often struggle when applied directly to production environments.
A standardized algorithm cannot solve every frontline problem.
Sany Heavy Industry faced this challenge when it introduced a general AI vision model for parts inspection. But variations in workshop lighting and the complexity of component textures produced too many false readings, making the system unsuitable for mass production.
However, the company did not give up. Its technical team worked in production workshops, collected large amounts of real-world data and repeatedly refined the algorithm to match the manufacturing process.
The result was a customized system that significantly improved inspection accuracy and efficiency. The mature solution was then replicated across production bases nationwide, turning AI from a single experiment into a scaled industrial application.
There”s a similar lesson in grassroots governance. Hidden risks are often discovered late, while manual inspections can be inefficient.
General-purpose large models may fail to identify detailed street-level problems such as road obstructions, cluttered public spaces or minor fire hazards.
When Beyondsoft developed smart community projects, it did not rely on an off-the-shelf solution. Its technical team followed frontline grid workers, collected real governance data, aligned algorithm rules with enforcement standards, and refined the system category by category.
After being adjusted in real-life scenarios, the intelligent inspection system became more stable and adaptable. It is now deployed in multiple cities, providing a replicable model for the digital transformation of urban governance.
To scale up AI innovation, China must strengthen collaboration between industry, universities, research institutes and users. This way, innovation will respond precisely to industrial needs.
The XCMG Group is a good example. Through a laboratory jointly established with universities, operational, fault and energy-consumption data from equipment across the country were synchronized with the research side in real time.
Once solutions were validated on prototypes, they were quickly deployed across tens of thousands of machines already in use.
This created a virtuous cycle of iterative improvement, scenario testing and large-scale deployment, significantly shortening the journey from research to application.
Capital is another key requirement for moving AI from pilot projects to widespread adoption.
Industrial AI often has lengthy development cycles, heavy investment and delayed returns. Such innovation depends on patient, long-term capital.
In the past, capital markets often chased the latest model concepts and laboratory breakthroughs, favoring sectors that promised immediate attention and faster monetization.
As a result, companies that worked to apply AI deeply in real economic scenarios often struggled to secure financing.
Aqrose Technology shows why patient capital matters. Focused on light manufacturing, the company serves small and medium-sized manufacturers in sectors such as hardware, home appliances and furniture, providing lightweight intelligent quality-inspection solutions. For this, it had to work extensively on the front line, collect a huge number of samples and develop standardized, cost-effective products.
As the practical value of its technology became evident, long-term industrial capital started flowing in. That allowed the company to refine its products and expand services to thousands of manufacturers. Industrial AI thus moved beyond isolated pilot projects and evolved into a large-scale application model.
Institutions and forward-looking policies are the foundation for sustainably scaling up AI innovation.
As AI becomes more deeply embedded in the economy and society, it will reshape production models, job structures and industry rules. At the same time, it will also raise new governance challenges involving data flows, algorithm compliance and risk management.
Without unified rules, AI applications can become fragmented and difficult to replicate. Futian district in Shenzhen, Guangdong province, faced these challenges during its AI-powered government service transformation.
By establishing standards for government AI applications, unifying rules for document review, compliance checks and data use, and simplifying approval procedures for intelligent government systems, it created a framework that could be replicated elsewhere. The experience demonstrates how institutional innovation can support technological innovation.
Global AI competition has already moved beyond isolated technological rivalry to a systematic contest of ecosystems, industrial integration and scaling capability.
While some developed countries lead in algorithms and models, China’s complete industrial system, vast market scenarios and strong industrial support give it natural advantages in adapting, testing and replicating AI technologies across sectors.
The Summer Davos’ focus on scaled-up innovation captures this broader shift.
In the AI era, technological breakthroughs are only the entry ticket. What determines long-term success is the ability to connect technology with real-world scenarios, adapt it to industrial environments and build innovation ecosystems.
Looking ahead, China should continue to strengthen basic research and protect the roots of technological innovation.
But it must also rely on high-quality industrial scenarios, rational long-term capital and sound institutional policies to open up the full chain from technological invention to industrial application.
Only when AI takes root in the real economy and empowers thousands of industries can it unleash the momentum of new quality productive forces.
The author is the founding director of the China Institute of New Economy.
The views don’t necessarily reflect those of China Daily.
If you have a specific expertise, or would like to share your thought about our stories, then send us your writings at opinion@chinadaily.com.cn, and comment@chinadaily.com.cn.
