On April 2, 2026, according to East Eight Zone time, the General Office of the Ministry of Industry and Information Technology officially released the "Notice on Carrying out Special Actions to Empower Small and Medium Enterprises Development with Inclusive Computing Power," placing the computing power shortcomings of small and medium enterprises on the policy agenda. The document proposes to rely on three major platforms: The SME Zone of China Computing Power Platform, The Computing Power Resource Connection Zone of China SME Service Network, and The National Computing Power Internet Service Node Zone, to optimize the supply structure of computing power and explore new models such as "computing power banks" and "computing power supermarkets." In reality, one end shows that China's overall computing power scale has ranked second globally for three consecutive years, presenting a façade of "abundance," while the other end reveals that the cost of computing power for small and medium enterprises is 40%-60% higher than that of larger enterprises, creating a situation where they struggle to keep up. In this structural contradiction, whether computing power will transition from a scarce resource to an infrastructure similar to water and electricity, or whether it will be further strengthened in a new round of centralization, has become the core of controversy under current policy narratives.

Strained Computing Power: Why Are Small and Medium Enterprises Left Out?

Superficially, China's total computing power scale has consistently ranked second globally for three years, with data centers and intelligent computing clusters densely established in multiple regions, fostering a perception within the industry that "overall computing power is not lacking." However, when the perspective shifts from macro to the enterprise level, especially for small and medium enterprises, the story changes completely: a substantial total resource does not mean that affordable, accessible, and sustainable computing power truly reaches long-tail users.

According to data disclosed in a relevant white paper for 2025, the average cost of computing power usage for small and medium enterprises is 40%-60% higher than that of larger enterprises. The reasons behind this are not complicated: large enterprises, leveraging their scale advantages and long-term contracts, can secure cloud resources and dedicated line cabinets at lower discounted rates; leading institutions may also build or deeply customize computing clusters, spreading out equipment depreciation and operational costs. In contrast, small and medium enterprises mainly have scattered, short-term, peak demand, possess weaker bargaining power, and can only endure standard prices or even overpricing, resulting in higher costs for equivalent computing power.

In practical scenarios, this cost gap manifests as an invisible barrier. Startups adjusting vertical models often find themselves deterred by high quotes while estimating training budgets, forced to revert to traditional algorithms or lower configuration solutions; manufacturing small and medium enterprises hoping to optimize supply chains and production processes through big data analysis often experience delays in digital upgrades due to the mismatch between computing power costs and project returns. AI training, batch simulation, log analysis, and localized search, which should become "standard capabilities," are begrudgingly relegated to "elective courses."

A deeper structural issue lies in the fact that computing power resources are highly concentrated in a few cloud vendors, large model companies, and leading internet enterprises. High-quality computing power—including next-generation GPU clusters, high-bandwidth low-latency networks, and nodes close to data sources—are often prioritized for major clients and personal business needs, leaving long-tail small and medium enterprises to “queue” at the tail end of resource allocation or being guided toward subpar performance and non-advantageous pricing combinations. While the total computing power footprint is expanding, the allocation mechanism reinforces the mismatch pattern of "the stronger the head, the harder for the tail."

Three Major Platforms Interlinking: How to Build a National-Level Computing Power Entry?

This special action takes the three major platform zones as the infrastructure for policy implementation: The SME Zone of China Computing Power Platform is responsible for aggregating various computing power supply and forming a unified display and access entry for small and medium enterprises; The Computing Power Resource Connection Zone of China SME Service Network focuses on matchmaking services, establishing a more efficient connection link between enterprise demands and specific computing power products and technical service providers; while The National Computing Power Internet Service Node Zone undertakes underlying network and node resources, ensuring cross-regional and cross-entity computing power scheduling capabilities. Although each has a different responsibility, all point to the same goal: to pull the computing power scattered among various parties "to a front desk," reorganizing it for small and medium enterprises.

Combined with the timeline proposed by the policy, it can be seen that this action is a further institutionalization and productized extension of the preliminary establishment of cross-regional scheduling capabilities of the national computing power network in 2025: the previous phase primarily addressed foundational issues such as "network connectivity and node interconnectivity," while this time it layers on application-level designs of "SME zones" and "unified entries." Through a combination of a unified front portal, intelligent scheduling in the middle layer, and a multi-party resource pool in the backend, the policy aims to transform the previously isolated and fragmented computing power islands into a public service interface friendly to small and medium enterprises.

In the market, the evaluation of the "first national-level computing power resource sharing policy" is a recognition of this architecture: past computing power sharing largely remained within the confines of a single cloud vendor and a single region's internal scheduling, while this time, the action initiated by the Ministry of Industry and Information Technology implies that computing power resource sharing has risen from an enterprise action to national-level rules and platforms, marking a milestone. It not only changes "who schedules computing power" but also transforms "based on what principles computing power is scheduled."

From the supply side, the direct effect of platformization is to promote the improvement of existing computing power utilization rates and passively level disparities between regions and entities. Through the cross-regional scheduling capacity of the national computing power network, nodes in the western regions, which enjoy low electricity prices and favorable natural conditions, can "remotely supply electricity" to regions in the east with dense computing power demands; small and medium cloud vendors and data center operators also have the opportunity to access more orders through the platform, reducing resource idleness. For small and medium enterprises, the formation mechanism of computing power prices is expected to shift from "passively accepting single quotes" to "multi-party competition and intelligent recommendations under a unified entry," gradually weakening the cost disparities caused by regional and scale differences through the policy design of "unified entry + intelligent scheduling layer."

From Water and Electricity to Computing Power: The Turn Toward Infrastructure Narratives

After the policy's release, some media commentators pointed out that this special action "marks that computing power resources are starting to align with infrastructure such as water and electricity." This is not a simple metaphor but a precise encapsulation of the policy thought: when computing power is written into public policy and coordinated through national-level networks and platforms, it is already viewed as a public production factor akin to land, energy, and communications, rather than merely a private technological asset of certain enterprises.

Looking back at the development trajectory of traditional infrastructures like water, electricity, and bandwidth in China, a roughly similar curve can be seen: early on, prices were high, and supply concentrated in a few regions and entities, making operational costs for enterprises enormous; as the country undertook unified planning, cross-regional transmission, and price mechanism reforms, resources gradually shifted from "scarcity" to "universal benefit," eventually sinking to the subsurface of industrial cost structures. Computing power is likely to follow a similar path: in the medium to long term, through the national computing power network and multi-party profit-sharing mechanisms, price volatility may decrease, availability may enhance, and it could become a foundational investment that enterprises can stabilize in their annual and project expectations.

For small and medium enterprises undergoing digitalization and AI transformation, the long-term impact of the infrastructureization of computing power far exceeds the immediate benefits. Once computing power becomes standardized and public utility-like, enterprises in the design phase can proactively embed more algorithms and data capabilities rather than being forced to "trim down" due to budget constraints post-factum. This would, in turn, change the product forms and competitive strategies of small and medium enterprises: traditional industries can experiment with low-cost intelligent production scheduling, predictive maintenance, and small model private domain applications, transitioning from "whether computing power can be used" to "how to use computing power well."

However, this path does not solely benefit small and medium enterprises; it presents a challenge to existing structures as well. Large leading companies have enjoyed significant benefits from self-built computing power and bulk procurement in the past: the convergence of technological and resource barriers makes it challenging for latecomers to bridge the resource end shortfall, even if they have algorithms and scenarios. When the state begins to reshape computing power supply with an infrastructural logic, the "absolute barriers" that large companies have on computing power will be partially diluted, shifting the discourse from "who controls more resources" to "who understands how to efficiently serve more users." This also implies that the game surrounding access rules, pricing models, and service priorities has just begun.

Computing Power Banks and Computing Power Supermarkets: Experimenting with Financial and Market Mechanisms

The repeatedly mentioned "strong>computing power banks” and “computing power supermarkets” in this special action essentially decompose computing power into more granular, configurable "asset units," enabling on-demand purchasing and matchmaking transactions through a unified platform. The computing power bank can be understood as a "resource asset manager" for both the supply and demand sides: responsible for packaging, decomposing, and registering computing power from large data centers and cloud vendors, as well as providing unified settlement and risk control; the computing power supermarket serves as the frontend interface for small and medium enterprises, displaying computing power products of various types and price ranges, allowing enterprises to balance specifications, prices, and quality of service as they would when selecting cloud services.

In a more forward-looking vision, once computing power is standardized and assetized, it will naturally lead to a series of explorations into "financialization of computing power": including pricing mechanisms for computing power based on varying performance and availability, differentiated rights of resource usage for different durations, and pledge financing based on long-term computing power quotas, among others. These explorations are currently still at the conceptual and pilot levels, and policy documents purposefully use cautious terms such as "exploration" to avoid creating overly strong short-term market expectations, but the direction is already clear—computing power is not only a resource that can be "sold" but can also become a configurable and tradable productive asset.

In a context where regulatory and risk boundaries are not yet completely clear, the policy emphasizes "exploration," reflecting a deliberately slowed pace. On one hand, it aims to prevent issues arising from excessive leverage and complex derivatives in the assetization process of computing power; on the other hand, it needs to leave space for negotiation and iteration regarding pricing rights, measurement standards, and contractual constraints as foundational institutions. For small and medium enterprises, if this model progresses steadily, its potential benefits include:

● Relief of cash flow pressure. Enterprises can purchase or lock in computing power more flexibly based on demand instead of investing a large sum upfront to acquire hardware or sign rigid long-term contracts.

● Increased predictability of budgets. Standardized products and pricing mechanisms allow for more precise estimations of computing power costs during project initiation stages, reducing risks associated with resource fluctuations.

● Lower costs of technical trial and error. In an environment where computing power can be rented, shared, or returned, small and medium enterprises can experiment with new models and business processes more boldly without worrying that a trial error will result in sunk costs.

Who Will Foot the Bill: Central Dividends, Local Pressures, and Large Enterprises' Concessions

The inclusive computing power special action does not spring forth from nowhere; it is built upon the foundation of the previous construction of the national computing power network. It is expected that the central government will continue to assume design responsibilities for top-level planning, network standards, and platform rules as "hard constraints," while also providing certain construction support for key regions and nodes through fiscal and policy tools; meanwhile, local governments need to assume more responsibilities at the implementation level, including site selection for data centers, energy support, and industrial park acceptances. Specific standards for subsidies and implementation timelines have yet to be announced; how costs will be shared and how to balance interests across different regions can only be summarized as "details pending."

On the enterprise side, there are both conflicting interests and collaboration opportunities among cloud vendors, large model companies, and local state-owned enterprises. Cloud vendors and large model companies hope to leverage national platforms for more resources and policy support to expand their business boundaries; however, they also inevitably harbor concerns about price pressures and resource concessions arising from "inclusiveness." Local state-owned enterprises often control key factors like energy, data centers, and land and hope to use computing power construction to drive local industrial upgrades, but they also need to find a balance between returns and inclusive demands. For inclusive computing power to truly take root, some parties must make temporary concessions in profitability and market share in exchange for a larger industrial cake.

In this multi-party negotiation, small and medium enterprises need to seek a voice through more organized means to avoid being passively "represented." Industry associations can voice their needs during demand surveys, standard-setting, and pilot project selection; industrial parks and clusters can enhance overall bargaining power through centralized purchasing and unified access, transforming individual scattered needs into "order scales" that can be recognized by platforms. For policymakers, how to capture the genuine needs of these small and medium enterprises in institutional design will directly determine whether inclusive computing power can truly be deemed as "inclusive."

It must be made clear that the implementation effects of policies will inevitably face time lags and regional disparities. The establishment of national-level platforms, enhancement of node capabilities, and adjustments to pricing mechanisms cannot immediately make all small and medium enterprises feel a significant drop in computing power costs. In areas with weaker foundational conditions, computing power anxiety may even see a temporary intensification during the transition period—as expectations are raised, but supply-side adjustments have not been completed. For market participants, maintaining calm between expectation and reality is a prerequisite for this policy window period.

From Documents to Implementation: How Should Small and Medium Enterprises Prepare Their "Boarding Spots"?

In summary, the inclusive computing power special action holds noticeable institutional potential for mitigating cost disparities and optimizing resource allocation: through national-level platforms and networks, part of the high-quality computing power originally concentrated at the top can be "sunk," opening more digitalization and AI paths for small and medium enterprises in the medium to long term. However, whether it can actually result in perceptible cost changes depends on a multitude of variables such as subsidy strength, node layout, and market competition. Current details on subsidies, pilot lists, and supporting policies remain undisclosed, and there exist significant uncertainties regarding the strength and pace of policy transmission.

In this context, small and medium enterprises do not need to wait for all details to be finalized before taking action; they can proactively do a few relatively certain things:

● Continuously monitor the construction progress of the three major platform zones, familiarizing themselves with interfaces, product classifications, and application processes in advance to reserve technical and organizational preparations for future access.

● Actively participate in various pilot applications and demand surveys, using channels like industry associations, parks, and industrial alliances to bring their computing power needs and pain points as early as possible into the stages of policy design and platform rule formulation.

● Internally review and optimize the structure of computing power usage, distinguishing between long-term stable demands and short-term experimental needs, evaluating which can be transitioned to the future "computing power bank/computing power supermarket" model, allowing for quick switching and magnifying policy dividends once new mechanisms are in place.

In the future 3-5 years of the computing power infrastructure process, small and medium enterprises are likely to transition from passively "receiving digitalization" to actively "shaping AI applications" within the industrial chain. As the barriers to computing power are gradually lowered, what truly determines competitiveness will no longer be "who grabs the GPU first," but rather "who understands their business context better and can transform inclusive computing power into unique products and services." For today’s small and medium enterprises, what needs to be done now is not to wait for a perfect policy list but to ensure they have reserved their "boarding spots" for the upcoming inclusive computing power train.

Join our community, let's discuss and become stronger together!
Official Telegram community: https://t.me/aicoincn
AiCoin Chinese Twitter: https://x.com/AiCoinzh

OKX Benefit Group: https://aicoin.com/link/chat?cid=l61eM4owQ
Binance Benefit Group: https://aicoin.com/link/chat?cid=ynr7d1P6Z

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。