Industry Insights | The Core Infrastructure of the Intelligent Era: Reunderstanding the Long-Term Value of the Computing Power Industry
文章免費5 天前
Within the research framework of industrial transformation and long-term capital that the Global Value Investment Association has long focused on, we consistently hold that investment opportunities capable of truly transcending cycles and reshaping the global economic landscape rarely stem from short-term market sentiment, but emerge from the most fundamental productivity revolutions in the evolution of human civilization.
Over the past two years, the new round of technological revolution triggered by AI large models is reshaping the global industrial system and the pricing logic of capital markets at an unprecedented speed. From North American tech giants continuously expanding AI capital expenditures, to the global semiconductor industry entering a new boom cycle; from the accelerated construction of ultra-large-scale data centers, to a new round of strategic competition among countries around high-end chips, advanced manufacturing and energy systems, artificial intelligence has gradually evolved from a single technological innovation into a global infrastructure revolution that will influence the next few decades.
Capital markets are often the first to reflect the trends of the times. In recent years, the market capitalizations of core enterprises in the global AI industrial chain, represented by NVIDIA, Google, TSMC, Broadcom, Micron and SK Hynix, have expanded rapidly and continuously. What lies behind this is not just market sentiment-driven growth, but a systematic revaluation by global capital of the "value of core infrastructure in the intelligent era". To date, the combined market capitalization of leading tech companies in the U.S. AI industrial chain alone accounts for an extremely high proportion of the global capital market. Whether it is chips, memory, advanced packaging, high-speed interconnects, data centers or power systems, capital is concentrating on AI underlying infrastructure with unprecedented intensity.
This round of global asset revaluation essentially reflects the market's renewed understanding of the underlying productivity of the future intelligent society. Among all the core production factors of the intelligent era, the core variable that truly determines the future industrial height and technological boundaries is computing power.
I. Computing Power Will Become the Core Infrastructure of the Intelligent Era
Throughout the history of human civilization, it is essentially a history of technological breakthroughs that continuously push the boundaries of human physiology, environment and cognition. In every genuine industrial revolution, those who master the core infrastructure of the era gain the initiative for the next phase of development.
In the agricultural era, iron tools and irrigation drove improvements in production efficiency; in the industrial era, steam engines and electricity reshaped the global manufacturing system; in the information era, the internet became the world's most important information infrastructure; and now we are on the cusp of the intelligent era.
From an industrial logic perspective, the three most fundamental elements of the intelligent era are data, electricity and computing power. Yet the core resource that ultimately determines the upper limit of intelligence remains computing power.
Data is more like the "raw material" of the new era. Today, the world generates massive amounts of data every day, including text, images, audio, industrial parameters, user behavior and real-time interaction information. But data does not automatically create value on its own; it can only be truly transformed into intelligent capabilities through model training and inferential computing.
Electricity is closer to the basic energy system of the intelligent era. With the continuous expansion of AI data centers, High Performance Computing (HPC) and ultra-large-scale AI clusters have placed higher demands on energy systems, and many regions around the world have begun to re-plan energy supply and power infrastructure around data centers.
Yet the core bridge that transforms "data" into "intelligence" remains computing power. In other words, data determines the breadth of intelligence, algorithms determine its efficiency, and computing power determines its upper limit. This logic establishes the fundamental position of computing power in the future industrial system.
Whether it is industrial intelligence, autonomous driving, humanoid robots, AI Agents, life sciences, digital twins, or cutting-edge fields such as future space exploration and quantum technology, their underlying operations all require support from continuously growing training and inference computing power. Moreover, demand for computing power has extremely strong long-term sustainability. As intelligence gradually permeates all areas of society, computing power is becoming an indispensable basic resource for the operation of the entire society, just like electricity in the past. Once intelligence enters an industry, the demand for computing power is often irreversible.
II. The Most Important Industrial Characteristic of the AI Era: Explosive Short-Term Growth in Computing Power Demand and Mismatched Supply Capacity
One of the most distinctive features of the AI era is that demand for computing power will show obvious explosive growth.
Traditional consumer industries usually have strong demand elasticity, and economic cycle fluctuations directly affect end-user demand. However, AI's demand for computing power is more characterized by persistence, incrementality and long-term nature. The reason is that the performance improvement of large models itself is highly dependent on the scale of computing resources. Whether it is the expansion of model parameters, the enhancement of multimodal capabilities, or the improvement of complex reasoning capabilities, all require support from higher-level training and inference computing power.
The current market understanding of AI computing power still mostly stays at the "training phase". However, from the perspective of industrial evolution, the demand on the inference side is likely to have a much larger scale in the future. As AI gradually enters the stage of large-scale commercial application, end-user interactions, enterprise-level AI applications, deployment of intelligent terminals and real-time Agent systems will all significantly increase inference computing power consumption. Future demand for computing power will not grow simply linearly, but is more likely to show an exponential expansion trend.
At the same time, there are obvious constraints on the supply side. Advanced process nodes continue to approach physical limits, and R&D costs for processes below 2nm are rising rapidly; Extreme Ultraviolet (EUV) lithography equipment has long-term limited supply; key links such as High Bandwidth Memory (HBM), advanced packaging and high-speed interconnects have long production expansion cycles; and the construction of ultra-large-scale data centers is further restricted by energy systems and power capacity. Therefore, the conflict between short-term computing power demand and supply is very likely to intensify suddenly.
From a medium to long-term perspective, the global computing power industry will most likely continue to maintain an industrial pattern of "high demand growth + high supply barriers", and this supply-demand structure often means that core links in the industrial chain have stronger long-term value support.
III. The Long-Term Investment Value of Underlying Infrastructure in the Intelligent Era Has Emerged
Historically, in every industrial revolution, the companies with the most long-term value are usually not those that ride short-term traffic and trends, but those that master core infrastructure and key production tools. At the current juncture, the development path of the global computing power industry has actually gradually become clear. Rather than predicting which model company will ultimately win, long-term capital needs to focus more on which companies truly possess irreplaceable underlying infrastructure capabilities in the intelligent era.
The global computing power industry has now formed high technical and capital barriers. Core links including GPU chips, HBM memory, advanced packaging, semiconductor equipment, high-speed interconnects, optical modules, switches, data center infrastructure and energy systems all require long-term technological accumulation, huge capital investment and a global industrial collaboration system to support. From a long-term perspective, such enterprises with industrial chain control, technical barriers and scale advantages are more likely to continuously obtain excess returns in the intelligent era.
IV. Global Value Investment Association’s View: Computing Power May Become the Core Asset Direction for the Next 10 to 20 Years
The essence of investment is the realization of cognition about long-term industrial trends and the direction of the times. For long-term capital, what truly matters is never short-term market fluctuations, but the ability to identify era-defining industrial trends and which enterprises can continuously create long-term value in the process of industrial restructuring.
Looking back at the past two hundred years of global industrial evolution, whether it was railways, electricity, oil or the internet, every technological revolution that truly changed the way society operates has ultimately given birth to new core infrastructure and reshaped the logic of global capital allocation. Today, computing power is gradually playing a similar role.
At present, the AI industry may still be in the early stages of the intelligent revolution, but the global industrial system's dependence on intelligence is continuously increasing. And the computing power system that supports all of this is gradually becoming an important foundation for the future global economy.
The Global Value Investment Association believes that what the capital market truly deserves long-term attention in the future is not just the AI concept itself, but the reconstruction of underlying infrastructure corresponding to the intelligent era and the resulting long-term industrial value revaluation.
From this perspective, computing power is not just a logic of the technology industry, but may represent one of the most important infrastructure investment directions globally for the next 10 to 20 years.
"Truly great investment is never about chasing short-term sentiment, but about identifying the core infrastructure that will reshape the future world in the early stages of an era's transformation."
