From the Digital Economy to the Generative AI Revolution

Since the 1990s, China has experienced rapid development of the digital economy, spanning more than three decades to date. According to statistics from the China Academy of Information and Communications Technology (CAICT), industries highly correlated with the digital economy now account for over 40% of China's macroeconomy, and this share has grown faster than GDP for more than ten consecutive years. This clearly demonstrates that the digital economy has become the most dynamic engine of innovation within the broader economy.

In the course of the digital economy's development, digital technologies have played a crucial supporting role. The previous generation of digital technologies is often summarized as "ABCD"—AI, Blockchain, Cloud computing, and big Data analytics. The most prominent contribution of these technologies was to migrate offline activities online, break through the constraints of physical space, move business operations to the cloud, and enable efficient utilization of computing power. Through the Internet of Things, they achieved remote reach; through big data analytics, they enabled efficient matching between people and information; and through automated information systems, they improved the efficiency of business execution.

Over the past few years, we have continuously promoted the integration of the digital and real economies, and many large enterprises have achieved notable results. However, for small and medium-sized enterprises (SMEs) and certain traditional industries, the "last mile" of digital transformation still presents significant obstacles. These obstacles manifest primarily in two ways:

First, the costs of deploying, developing, and maintaining digital systems remain prohibitively high.

Second, the high precision demanded by digitalization creates an adaptation mismatch with "last mile" business scenarios. Frontline workers, sales personnel, and end users often encounter significant barriers when feeding information into the system.

Take our own NSD at Peking University as an example. We have been promoting information technology development for many years, but overall progress over the past decade has been less than ideal. Despite the school's modest size, its business scenarios are complex. Frontline faculty struggle to clearly articulate their specific needs for information systems. By the time a service provider completes development according to the stated requirements, those requirements have often already changed—and the resulting need for secondary development incurs substantial costs. These real-world challenges have created numerous obstacles for the digitalization process.

Over the past few years, my research focus has gradually shifted from the digital economy and digital finance to the field of AI. In particular, with the launch of ChatGPT in late 2022 and the release of DeepSeek in China in early 2024, we have officially entered the era of the generative AI revolution.

Four Breakthroughs Brought by AI

What are the fundamental differences between generative AI and the previous generation of digital technologies? Experts in the AI field can explain this at the technical level; I, from a functional perspective, summarize four key breakthroughs:

The First Breakthrough: The Ability to Creatively Generate Content

The core of previous technologies was precise matching and optimization-based decision-making. For example, AlphaGo playing Go was about finding the optimal move within a given board position—so earlier technologies excelled at precision but lacked creativity. Creativity was once considered a uniquely human ability. Today, however, generative AI has demonstrated a degree of creativity. Of course, this creativity comes with corresponding costs, namely the problem of model hallucination.

The Second Breakthrough: Transformed Human-Computer Interaction

This is critically important. Since the advent of computers and information technology, human-computer interaction has been a persistent challenge. From assembly language to high-level languages, from the Windows operating system to various programming languages, to direct a computer to accomplish a task, humans had to interact using precise languages that the computer could understand. In this model, the computer occupied the dominant position, and humans had to adapt to the computer's logic. The emergence of generative large language models has fundamentally transformed this interaction paradigm. Whether using Doubao, Gemini, or GPT, we can now issue instructions to AI using natural human language. This change has dramatically lowered the barrier to human-computer interaction and greatly improved the user experience. Its impact is especially significant for two groups: first, business managers who previously had to rely on technical staff to interact with systems due to unfamiliarity with system operations; and second, frontline business personnel who generally lacked programming skills and similarly depended on predetermined systems to interact with machines. Now, the natural-language-based interaction model will greatly facilitate the connection of frontline information, knowledge, and experience to AI systems, particularly enabling an end-to-end pipeline for data collection, organization, and analysis.

The Third Breakthrough: Rapid Advances in Multimodal Technology

Previous information technologies were primarily based on data and text content, with some limited ability to process images, but very weak and prohibitively expensive capabilities for video processing. In recent years, AI's multimodal processing capabilities have improved substantially. Last July and August, I used AI technology to create a personal vocal performance piece, "Song of AI Voyage," at a total cost of less than 100 yuan, with strong dissemination results and click-through rates. This year, video generation capabilities have been further upgraded, and AI short dramas and AI comic dramas have already developed into an entirely new industry track.

The Fourth Breakthrough: Significantly Enhanced Logical Reasoning

In the early days of large language models, logical reasoning capabilities were still weak. However, with the emergence and continuous optimization of reasoning models, large models can now complete complex tasks and, on certain scientific tasks, have reached the level of top-tier learners.

All four of these capabilities represent critical breakthroughs achieved only in the past few years.

Since the explosion of generative large language models, massive amounts of capital have accelerated into the AI field. According to incomplete statistics, over half of new financing in Silicon Valley has flowed to AI-related startups, and capital expenditure by global technology companies has grown substantially. At the same time, the user base for AI applications continues to expand at high speed, large model performance keeps improving, and the upgrade cycle far outpaces the rhythm of Moore's Law—chip performance roughly doubles every 18 months, while the capability upgrade cycle for large models is only 3 to 6 months. Meanwhile, the costs of model training and usage continue to decline, and AI-related infrastructure keeps improving.

Looking at the trajectory of technological evolution, this round of AI progress has achieved a triple leap: from decision-based AI to generative AI, from the era of small models to the era of large models, and from specialized AI technology to general-purpose AI capabilities. Previously, different AI applications—natural language processing, facial recognition, big data analytics—each required dedicated models and data support. Today's large models, however, are like all-around athletes, possessing cross-domain general capabilities. This is why people are now discussing whether this round of AI technology, like the steam engine of the First Industrial Revolution, the electric motor and generator of the Second, and the computer and internet of the Third, will bring about massive productivity gains across all industries and thereby catalyze a Fourth Industrial Revolution.

In the era of intelligence, a pivotal change is that "intelligence" as a new factor of production has achieved large-scale, low-cost supply. Previously, for a company to apply intelligent capabilities in its business activities, it needed to pay very high costs and meet numerous conditions. Today, intelligent capabilities can be deployed at scale and at low cost.

New Industries, New Business, and New Paradigms Driven by AI

What industrial transformations has the AI technology revolution brought about?

1. The Development Opportunities of the AI Core Industry Chain

The core AI industry comprises three key elements: computing power, algorithms, and data. Currently, both the AI user base and token consumption are in a phase of exponential growth, directly driving the AI core industry—especially upstream computing power infrastructure—to achieve explosive growth.

Observing the capital markets, one finds that AI hardware companies, particularly those in high-tech sectors facing production bottlenecks and "chokepoint" risks, have actually seen capital market performance far exceeding that of AI application companies. The underlying logic is this: AI applications are growing exponentially, but not all applications are monetized and generate revenue. The computing power demand that supports these applications, however, is real and rigid. Unlike the expansion on the application side, the expansion of computing power is typically linear. When linearly growing computing power supply meets exponentially growing application demand, it drives up the prices of related products, which in turn drives substantial increases in capital market valuations.

Therefore, the development of the computing power industry manifests in two aspects: first, the increase in supply volume; and second, the need for traditional computing power infrastructure to undergo technological restructuring to meet AI computing demands. From chips, servers, and communications to storage and cloud services, the entire industry chain needs to upgrade in the direction of adapting to AI computing power requirements.

In this process, two major AI industry ecosystems have essentially formed globally—one centered in China and one in the United States—with both sides emphasizing self-reliance and supply chain security. Most other countries must choose to integrate into one of these ecosystems, or embed themselves into a specific segment of the value chain. South Korea is a prime example. It is arguably the country most deeply affected by AI globally, with approximately half of its GDP directly related to AI. Samsung and SK Hynix together account for over 50% of the Korean stock market's total capitalization. Korea successfully positioned itself in the chip and high-bandwidth memory track, and the stock price gains of related companies have been remarkable.

This process also contains enormous opportunities for domestic substitution. The state must ensure self-sufficient supply capacity in "chokepoint" areas and cultivate domestic alternative solutions. This is like horse racing: we need to nurture multiple competitive market entities, provide them with sufficient orders and capital support, and ultimately, whichever company emerges victorious will ensure supply chain security.

2. The Broader Opportunity of "AI + All Industries"

The enhancement of AI capabilities, particularly the four breakthroughs described above, has overcome the bottlenecks that previously limited the real-world deployment of digital technologies. This enables intelligent capabilities to drive innovation in products, services, and business models. Like the internet, digital technology, and big data before it, AI will become the foundational platform and innovation engine for future business development, reshaping business models across all industries.

3. The Opportunity of "AI + People"

The profound impact of AI on individuals is the core reason this round of AI revolution has drawn such intense global attention. Take a recent example: it was graduation season at universities, and every institution invited guest speakers for commencement addresses. I noticed that virtually every speaker's address mentioned AI—some framed it as the backdrop of our era, some explored its impact on academic disciplines and education, and others shared their perspectives on its implications for personal development and how to respond. AI has become an unavoidable topic of our time.

An interesting phenomenon is that several top American universities invited AI industry executives and founders to deliver commencement speeches, only to face open boycotts from some students. Behind this phenomenon lies the widespread anxiety among the American public—and even some elite students—about the enormous disruptions brought by AI. By contrast, China's overall attitude toward AI is more positive, with a greater inclination toward using AI well to create a better future.

The true impact of this round of AI revolution is its impact on people—and it affects virtually everyone.

On one hand, AI can serve as a "compressor" of individual capability differences. For instance, when I assign research tasks to students, the quality of their work used to vary greatly—the best students produced excellent work, while those with weaker foundations might fail. But now, submitted work consistently reaches a passing standard of 60 points or above, with AI rapidly closing the gap in foundational capabilities.

On the other hand, AI is also an "amplifier" of individual capabilities. If the instructor does not set strict requirements, the vast majority of students submit work that appears adequate but lacks genuine value. Only a small number of students integrate their own thinking and elevate their work to 90 points or above—and the gap between the two groups is actually widened.

The AI era will bring a "switch": some people will turn off the switch of independent deep thinking, delegating more tasks to AI. Others will make good use of AI tools, propelling themselves into a phase of deep learning and dramatically enhanced capabilities. The capability gap between these two types of people will be significantly amplified.

Extending the perspective from individuals to teams and organizations, one finds that AI similarly amplifies differences between teams and organizations. At the macro level, AI—like every previous technological advancement—will on one hand replace certain jobs and on the other create new employment opportunities. Past digitalization technologies primarily replaced blue-collar positions, while this round of AI may replace a large number of white-collar positions, including many roles for college graduates and even high-paying professional jobs. At the same time, however, AI will also spawn new positions—for example, the Chief Intelligence Officer is a new role that has emerged in response to this trend, and it demands even higher capabilities from practitioners.

Whether AI ultimately empowers or replaces humans, whether it leads to mass unemployment or propels humanity to a higher level of development, depends both on how humanity responds and on national policy directions. I am personally relatively optimistic on this front. I believe humanity possesses sufficient wisdom, and that decision-makers have sufficient strategic resolve, to make AI a tool that empowers human development.

Already, AI has brought about a comprehensive restructuring of many professions and skill systems. Market demand for some occupations has declined sharply. In many other professions, those who do not master AI tools will find it difficult to meet job requirements. We have entered the era of human-machine collaboration. AI has already become an assistant, a tool, and a collaborative partner that works alongside us.

How to manage human-machine collaboration is a question that every manager and every AI architect must consider.

The Connotation and Characteristics of the Smart Economy

At the national policy level, in July 2025, the State Council issued the Opinions on Deepening the Implementation of the "AI+" Initiative. This policy also reflects a significant difference between the AI development strategies of China and the United States: the U.S. treats AI as a core pillar industry with an extremely high share of GDP, while China places greater emphasis on using AI to drive efficiency improvements and business growth across all industries, promoting the intelligent upgrading of all factors and all sectors.

In the government work report delivered at the beginning of 2026, Premier Li Qiang specifically introduced the concept of "forging new forms of smart economy." This formulation is relatively novel. In the past, we talked about the "digital economy"; now we have begun speaking of the "smart economy." This marks the opening of a new stage of development.

How should we understand the smart economy and its relationship with the digital economy? I believe it can be interpreted from two dimensions. First, at the current stage, the smart economy can still be subsumed within the category of the digital economy, as its most cutting-edge and most active component. Second, from a longer-term perspective, the smart economy may represent a new economic form that succeeds the digital economy.

The smart economy does not yet have a single, strict, unified definition. Drawing on relevant policy documents, I have compiled a reference definition: the smart economy is a new economic form driven by next-generation AI technology as its core engine, with algorithms, computing power, and data as its key elements, and the "AI+" initiative as its implementation pathway, driving systemic transformation of production methods, industrial ecosystems, and value creation.

In the era of the digital economy, the metrics and value measures were traffic, attention, and user scale. In the era of the smart economy, the unit of measurement and value measure has shifted to the token—the basic unit of information processed by AI.

As shown in Figure 1, the smart economy and the digital economy differ markedly across several dimensions, including core technological characteristics, key elements, core technologies, measurement standards, sources of value creation, organizational forms, typical scenarios, and human-machine relationships. Specifically, the smart economy is characterized by AIGC, multimodal capabilities, logical reasoning, and human-machine collaboration; it is supported by the synergy of computing power, algorithms, and data; it relies on large models, computing power, and electricity as core technologies; it uses the token as its unit of measurement; and it achieves value creation through intelligence-driven processes. Its organizational forms include AI-native organizations, human-machine collaboration, and the Internet of Agents. Typical application scenarios encompass programming, office assistants, video entertainment, embodied AI, autonomous driving, and intelligent manufacturing. Human-machine relationships follow a collaborative co-creation model, encompassing AI assistants, human-machine collaboration, and digital employees.

Growth Opportunities and Practical Pathways in the Smart Economy

This year marks the inaugural year of the 15th Five-Year Plan. Comparing the 14th and 15th Five-Year Plans, the biggest variable is the entry into an entirely new stage of AI technology development.

In my view, the grandest commercial and growth opportunity during the 15th Five-Year Plan period is the deep integration of AI technology into the real economy, its embedding into business processes, and its empowerment of enterprise management and organizational transformation—creating value through these practical applications. But this process is by no means easy, because there is a fundamental difference between a large model possessing powerful capabilities on the one hand, and actually entering an enterprise, integrating into business scenarios, and creating value on the other.

This is akin to the cognitive gap between buyer and seller, or between client and vendor—there exists a massive implementation chasm between the two. Bridging this chasm during the 15th Five-Year Plan period will contain enormous growth opportunities. For large enterprises, the level of AI application will become the core determinant of industry positioning and competitive advantage. For SMEs, effectively leveraging AI can reduce operating costs, enhance organizational resilience, and help navigate uncertainty—it is a survival imperative. For managers, mastering AI application capabilities is a required course that must be completed.

Key Insight

The CIO's acronym remains the same, but the role must evolve from Chief Information Officer to Chief Intelligence Officer—Information becomes Intelligence. The most critical upgrade is deepening the understanding of business logic, especially the underlying logic of AI-driven business transformation.

CIOs undoubtedly possess solid technical foundations. However, I believe the important lesson CIOs need to learn is how to persuade enterprise decision-makers, collaborate with marketing and HR departments, and engage deeply with customers. The English abbreviation remains "CIO," but in essence it must upgrade from Chief Information Officer to Chief Intelligence Officer—Information becomes Intelligence. The most critical aspect is deepening the understanding of business logic, especially the underlying logic of AI-driven business transformation.

There is a widely circulated saying lately: "Every industry is worth redoing with AI." I find this formulation slightly absolute. I prefer an alternative phrasing: every industry is worth considering whether it should be redone with AI. Enterprises first need to assess whether their industry requires AI restructuring, and when to pursue it. Some industries can take the lead; others can proceed steadily.

From a competitive perspective, AI will determine a company's core competitiveness and industry positioning. For industries that are farther from digital technology—such as food and beverage, offline manufacturing, and other traditional industries, as well as those dominated by SMEs—AI services may not be provided by internet giants, but rather by core supply chain leaders within the industry. This is because industrial AI services require deep insight into the industry's business and mastery of domain-specific data, which major tech companies often cannot comprehensively cover across all niche areas.

Six Favorable Factors for Industry Development in 2026

In my view, industry development in 2026 benefits from the following favorable factors:

First, capital market enthusiasm for AI remains high. For instance, after MiniMax's IPO, its valuation exceeded one trillion yuan, and a company's AI strategy directly influences its capital market valuation. Therefore, a key function of the CIO-to-Chief Intelligence Officer transition is to help enterprise decision-makers articulate overall strategy from an AI perspective, clarify AI's positioning within the company's development strategy, and chart a concrete pathway for the enterprise to embrace AI.

Second, policies are working in concert. Whether in AI infrastructure construction, application scenario expansion, intelligent terminal procurement, or the development of a high-quality data factor market, relevant policies have been deployed and will be progressively implemented during the 15th Five-Year Plan period.

Third, large model capabilities have further strengthened, and AI agents have begun to see real-world deployment. Deploying agents into real scenarios carries non-trivial risks. Standardized agent practices—exemplified by initiatives like the "Lobster Battle" benchmark—their most important significance lies in driving the formation of industry standards, establishing safety boundaries, and enabling agents to deliver their capabilities while keeping risks controllable.

Fourth, tokens have begun achieving genuine commercial monetization—a highly symbolic milestone. A long-standing issue in the development of China's internet and digital economy has been users' reluctance to pay. In the smart economy era, the popularization of the token-based payment model is a very positive change. Only demand that people are willing to pay for is genuine demand.

Fifth, super individuals and AI-native organizations continue to emerge. AI has spawned many "super individuals"—those who skillfully use AI can independently accomplish work that previously required a dozen or more people, creating substantial commercial value directly through platforms. AI-native organizations are new entities that, from their founding, use large models as their technological foundation and human-machine collaboration as their core operating model. Their business, management, and culture are inherently aligned with the smart economy era.

Sixth, a growing number of enterprises are substantively driving AI implementation. The CIO's role in enterprise strategic positioning will become increasingly important. In this process, the CIO needs to balance multiple dimensions—business, strategy, organization, and talent—to drive AI to deliver tangible results.

Finally, let me close with a passage I wrote: "The principal contradiction in the AI world today is between the growing demand of the broad masses for the advanced productive forces of AI, and the gap among the business community in understanding AI, knowing how to deploy it, knowing how to use it, daring to apply it, and knowing where to begin." In the process of resolving this contradiction, enterprise CIOs have a vital role to play.