How can the insurance industry make early moves in the trillion-yuan AI infrastructure track?
Artificial Intelligence (AI) is entering a new cycle of capital expenditure. In its latest report, Goldman Sachs points out that the focus of AI investment has begun to shift from model R&D to the broader real economy. While computing power, power supply and data centers continue to be constructed at a high speed, AI is accelerating its penetration into real-world scenarios such as manufacturing, energy, logistics, national defense, life sciences and robotics. Goldman Sachs forecasts that from 2026 to 2031, global AI capital expenditure centered on computing, data centers and power will reach approximately $7.6 trillion.
China is also rapidly entering this round of AI infrastructure construction cycle. In early 2026, the daily average Token call volume of China's AI large models has exceeded 140 trillion, representing a roughly 1000-fold increase compared to two years ago; there are more than 500 intelligent computing center projects in operation, under construction or in planning across the country, with the intelligent computing power scale reaching 1.882 million P (P stands for one quadrillion floating-point operations per second) (FP16), ranking second in the world; during the "15th Five-Year Plan" period, the investment scale of China's computing power infrastructure is expected to be around 2 trillion yuan, and a new digital infrastructure system covering intelligent computing centers, GPU (Graphics Processing Unit) clusters and computing power networks is taking shape at an accelerated pace.
As artificial intelligence enters the era of capital expenditure, a previously non-existent AI infrastructure asset pool is also rapidly forming. Accompanied by trillions of yuan in infrastructure investment, not only new industrial opportunities have emerged, but a large amount of high-value assets are also continuously accumulating. For the insurance industry, a new risk pool is taking shape, and the corresponding risk protection system remains to be improved.
AI Enters the Era of Capital Expenditure, A New Asset Pool Is Taking Shape
Is China forming a sufficiently large AI infrastructure asset pool? Is this asset pool large enough to change the long-established risk structure of the insurance industry?
The answer is gradually becoming clear.
As the application scope of large models continues to expand, the call scale of artificial intelligence models is growing rapidly. Data shows that the daily average Token call volume of domestic AI large models has increased by approximately 1000-fold within two years, rising from 140 billion in early 2024 to 140 trillion in March 2026. Artificial intelligence is evolving from a technological experiment of a few tech enterprises to a stage of large-scale commercial application in industries such as finance, manufacturing, energy and healthcare.
Every time a technology moves from the experimental stage to industrialization, it is accompanied by the launch of an infrastructure investment cycle.
In the past few years, the competition in the artificial intelligence industry has mainly focused on model capabilities, with the parameter scale of large models continuously expanding and algorithm capabilities making continuous breakthroughs. However, as generative artificial intelligence enters the stage of large-scale application, industrial competition is changing: artificial intelligence is shifting from "model competition" to "infrastructure competition".
The key resources supporting AI development are shifting from software capabilities to a more capital-intensive infrastructure system. Large model training, inference services and enterprise intelligent applications all require continuous investment of massive computing resources, and computing power is becoming a new type of infrastructure in the artificial intelligence era, similar to power, communication and transportation. According to relevant data, as of March 2026, China's intelligent computing power scale reached 1.882 million P (FP16), 2.5 times higher than the same period last year; the scale of China's computing power market is expected to reach 835.1 billion yuan in 2025, and is likely to exceed 1 trillion yuan in 2026. Meanwhile, there are already more than 500 intelligent computing center projects in operation, under construction or in planning across the country, and 10,000-GPU-level intelligent computing clusters are constantly increasing.
Different from the rapid expansion of the past internet industry relying on software, platforms and traffic, the development of artificial intelligence has more distinct capital-intensive characteristics. A large-scale intelligent computing center is not a simple stack of servers, but a complex infrastructure system composed of buildings, power supply systems, high-performance GPU servers, cooling systems, network equipment and security facilities.
From a global perspective, the scale of AI infrastructure investment is reaching an unprecedented level. Relevant research from Swiss Re points out that the development of artificial intelligence is driving a new capital expenditure cycle and creating a large number of previously non-existent insurable assets. The construction cost of some large data centers has reached billions of dollars, and after deploying high-value GPU equipment, the overall investment scale may further increase, with super-large projects even exceeding $20 billion.
This means that AI infrastructure is forming a new high-value asset pool. From the perspective of the insurance industry, this change is of great significance.
In the past few decades, property insurance has mainly focused on risk assets formed in the industrial era, including industrial plants, production equipment, commercial buildings and energy and transportation facilities. The artificial intelligence era is creating new objects of risk exposure, including intelligent computing centers, GPU clusters, computing power networks, and the digital infrastructure system formed around the operation of artificial intelligence.
As new-type infrastructure projects with individual investment scale reaching billions or even tens of billions of dollars continue to increase, it not only brings opportunities to the technology industry, but also means that a new risk protection market is taking shape.
Super Computing Power Assets Are Creating Super Risks
The large-scale construction of artificial intelligence infrastructure is creating a new high-value asset pool. However, for the insurance industry, what truly deserves attention is not only the rapid growth of asset scale, but also the profound changes in the risk patterns behind these assets.
Because there is a fundamental difference between AI infrastructure and traditional industrial assets: its value is not only reflected in the equipment itself, but in the ability to continuously provide computing power.
In the past few decades, property insurance has mainly centered on "physical asset loss". When a fire breaks out in an industrial plant, insurance focuses on the losses of buildings, production equipment and inventory; when an accident occurs in an energy facility, insurance focuses on equipment damage and engineering repair costs. The risk logic of traditional property insurance is essentially built around the value of fixed assets.
But for artificial intelligence infrastructure, risks are changing. The most important assets of an intelligent computing center are not just the GPU servers deployed in it, but the computing power production capacity formed by these devices together. Once a critical system fails, even without severe physical damage, it may cause huge economic losses.
This means that the problems faced by insurance in the AI era are shifting from "what to do when equipment is damaged" to "what to do when the computing power production system stops running".
AI infrastructure may face multiple risks at the same time, including asset loss, business interruption, and the integration of physical risks and digital risks. An intelligent computing center is composed of multiple complex systems such as high-value GPU servers, power supply systems, cooling equipment and network facilities. A failure in any critical link may affect the overall operation.
Compared with traditional industrial assets, AI infrastructure has higher technical complexity and system relevance. The risk is not a single equipment failure, but may involve the entire infrastructure system. At the same time, for enterprises relying on computing power services, the real loss is not only the equipment maintenance cost, but also the business impacts caused by the interruption of computing power services, such as delays in model training, suspension of artificial intelligence applications, interruption of customer services and performance risks of contracts.
In addition, the intelligent computing centers in the artificial intelligence era carry not only servers and data, but also core AI applications of enterprises, model training tasks and a large number of commercial services. Cybersecurity attacks, data breaches, ransomware and supply chain security risks are also becoming important factors affecting the operation of computing power infrastructure.
Especially for large-scale intelligent computing centers, the impact of cyber risks may far exceed that of traditional enterprise information security incidents. A cyber attack may lead to the interruption of computing power services, affect the progress of model training, and even cause the suspension of business of a large number of enterprise customers. Therefore, the insurance demand formed around AI infrastructure in the future will no longer fall into the category of a single insurance type, but needs to integrate multiple dimensions such as property loss, business interruption, cybersecurity, third-party liability and supply chain risks.
At the same time, the rapid development of AI infrastructure is also challenging the long-established underwriting and pricing capabilities of the insurance industry. When the investment scale of a single intelligent computing center reaches billions or even tens of billions of dollars, the risk diversification mechanism in the traditional insurance model will face new challenges.
For insurance companies, what they need to answer in the future is not just "whether to underwrite", but more importantly, how to accurately assess risks: How to quantify the risks of large-scale computing centers? How to price the combination of physical risks, cyber risks and business interruption risks? How to control the overall exposure of the industry when multiple large-scale projects encounter risks at the same time?
These issues mean that AI infrastructure insurance cannot simply copy the traditional property insurance model, but needs to establish a new risk assessment system. Insurance companies need to combine engineering risk management, cybersecurity capabilities, data analysis models and reinsurance mechanisms to improve their ability to identify and manage the risks of complex digital infrastructure.
From Risk Protection to Risk Management: Computing Power Insurance Reconstructs a New Ecosystem
AI infrastructure is creating a new high-value asset pool. But for the insurance industry, the real challenge is not just the emergence of market demand, but how to understand and manage a new type of risk that did not exist in the past.
The traditional insurance system is built on the accumulation of long-term risk data, and judges the probability of risk occurrence, loss degree and reasonable price through historical loss data. However, for intelligent computing centers, GPU clusters and computing power networks, many risks are still in the early stage, and the insurance industry is facing a new risk field of "high value, low historical data".
This makes computing power insurance first face the challenges of risk identification and pricing capabilities. Although large-scale intelligent computing centers have asset attributes similar to traditional infrastructure, their risk structure is more complex: on the one hand, physical risks such as equipment failures, power supply problems and cooling system failures are interrelated; on the other hand, digital risks such as cyber attacks, data security issues and business interruptions are further integrated with traditional property risks. Therefore, the traditional risk assessment model based on single asset loss is increasingly unable to adapt to the complex risks of AI infrastructure.
For insurance companies, the development direction of computing power insurance is not simply adding a new insurance product, but promoting the transformation of the risk management model. In the past, insurance mainly assumed the function of financial compensation after accidents; but for intelligent computing centers with an investment scale of billions of dollars, what enterprises need more is the risk management capability throughout the whole process of construction, operation and service. Insurance companies need to move risk management forward from post-accident indemnification through risk assessment, continuous monitoring, accident response and loss control.
This trend has already been reflected in the field of cybersecurity insurance. Taking the "prevention + insurance" integrated model explored by Yuanbao Technology as an example, in the face of the risks of digital infrastructure such as smart cities, cloud computing and big data, traditional information security protection can no longer cover all demands. Insurance institutions need to combine professional technological capabilities to identify risks before underwriting, continuously monitor risks during operation, and provide incident investigation, loss assessment and recovery support after accidents, so as to improve the overall risk governance capability.
In the artificial intelligence era, what insurance faces is not just a new business opportunity, but a new type of risk asset. In the future, the insurance system formed around AI infrastructure will no longer be limited to a single insurance type, but needs to integrate multi-dimensional protections such as property loss, engineering risks, business interruption, cybersecurity and third-party liability. At the same time, the insurance industry needs to gradually establish a risk data system and pricing model for AI infrastructure. Whoever can take the lead in accumulating the operation data, risk cases and loss experience of computing power infrastructure is more likely to gain the pricing power in this emerging risk market.
From Risk Protection to Long-Term Capital: The Insurance Industry Unlocks a New Role in the AI Era
The development of artificial intelligence infrastructure not only creates new insurance demands, but also brings new asset allocation opportunities. In the past few decades, with the attribute of long-term liabilities and stable capital sources, insurance funds have continuously participated in the investment of long-term assets such as transportation, energy and new-type infrastructure. As the AI industry enters the stage of infrastructure construction, new-type assets such as intelligent computing centers, data centers and computing power networks may also become important directions for long-term capital attention in the future.
AI infrastructure is characterized by a long investment cycle, large capital input and strong industrial driving effect, which is highly compatible with the characteristics of long-term investment and value investment of insurance funds. In recent years, insurance funds have become one of the important long-term capital sources for the ecological development of the artificial intelligence industry: on the one hand, insurance funds participate in the development of upstream and downstream enterprises in the AI industry chain through equity investment, industrial funds and other methods, providing long-term capital support for the construction of artificial intelligence infrastructure and the improvement of the industrial ecosystem.
On the other hand, as the scale of new-type assets such as computing centers and data infrastructure grows rapidly, the insurance industry is also providing risk protection for the construction of artificial intelligence infrastructure through innovative products such as computing power insurance, data center insurance and cybersecurity insurance. Insurance funds are gradually expanding from traditional financial investors to long-term capital partners for the development of the artificial intelligence industry.
In this sense, what AI brings to the insurance industry is not just a new type of insurance product opportunity, but development opportunities at different stages on both the asset side and the liability side. Insurance funds have become important participants in the investment of the artificial intelligence industry chain, supporting the development of AI infrastructure through long-term capital; at the same time, the new-type risk protection around AI infrastructure is still in the exploratory stage, and the insurance industry needs to further improve its risk identification and protection capabilities.
In the future, as computing power becomes an important means of production in the digital economy, whether the insurance industry can establish risk management capabilities matching the development of AI infrastructure will become an important window to observe how the insurance industry serves the development of new quality productive forces.
This article is from the WeChat Official Account "Caijing Mayflower" (ID: Caijing-MayFlower), written by Wang Yan, and published with authorization from 36Kr.