On September 4, the Ministry of Industry and Information Technology officially issued the "AI SME Entrepreneurship Support Plan (2026–2028)". The core message of this document is very clear: the country must systematically lower the threshold for AI entrepreneurship, so that "one person + AI" can also become a company.

The three-year goals are very specific: to cultivate more than 10,000 new technology- and innovative SMEs, surpass 2,000 "Little Giants" specialized, refined, distinctive, and innovative, and produce a batch of gazelle and unicorn enterprises.

At the same time, it will build 10 high-standard technology enterprise incubators, 10 national SME public service demonstration platforms, and 10 national-level SME characteristic industrial clusters.

In short: this isn't just throwing money around, it's about building an ecosystem. Next, Kingtech will help you break down the underlying logic and investment implications of this document.

|Why now? Because "one-person companies" are no longer just jokes

The background of this document is worth careful consideration. For the first time, the MIIT explicitly mentioned new types of business entities such as "One-Member Companies" (OPC) and super individuals in the document, and encouraged localities to provide "inclusive support."

What does this mean? The leap in large model capabilities has fundamentally changed individual productivity—with AI tools, a single person can complete the entire business operation from topic selection, R&D, customer service to finance. Content entrepreneurs use AI to complete the entire process from topic selection to publishing, while cross-border sellers rely on AI to handle multilingual customer service and product analysis. Annual revenue of millions per person is no longer just a myth.

The Small and Medium Enterprise Administration of the Ministry of Industry and Information Technology also admitted that this plan is designed to "adapt to technological evolution and changes in entrepreneurial forms," while also "breaking the constraints of elements in the startup phase, effectively lowering the threshold for entrepreneurship and the cost of trial and error."

Translated into plain language: the form of entrepreneurship has changed, and policies must keep up.

|The four pillars: computing power, data, scenarios, and capital—none can be missing

The document focuses on 15 key tasks across 4 areas. I will help you distill them into four key terms:

Computing Power — From "Affordable" to "Affordable"

Relying on the China Computing Power Platform, it promotes pooled integration and flexible allocation of computing resources, providing startups with low-cost, inclusive computing power.

Previously, the Ministry of Industry and Information Technology launched the "Special Action for Inclusive Computing Power Empowerment for SME Development," proposing innovative models such as "Computing Power Bank" and "Computing Power Supermarket"—allowing enterprises to flexibly purchase computing power based on "cardtime" and "verification time" as if buying daily necessities. This is a tangible cost reduction for startups.

Data—From "Unavailable" to "Orderly Opening"

Focusing on fields such as intelligent manufacturing, materials R&D, and biomedicine, support leading enterprises in forming data cooperation consortia with small and medium-sized enterprises, take the lead in building high-quality industry datasets, and orderly open them to startups. At the same time, it cultivates data service enterprises to provide data consulting, governance, annotation, and other services to reduce data acquisition costs.

Scenarios—From "Can't Find You" to "Chain Owners Take You Playing Together"

Relying on the "Hundred Events, Ten Thousand Enterprises" large, medium, and small enterprise integration and matchmaking activities, guide chain leaders to open up to startups for experimentation, demonstration applications, and large-scale commercial applications.

At the same time, local governments are supported in selecting "AI entrepreneurship scenario partners" to provide startups with full-chain services including scenario collection, demand analysis, pilot validation, and promotion application.

Capital—From "No One Invests" to "Patient Capital"

The document clearly proposes leveraging the guiding roles of the National SME Development Fund, the National Artificial Intelligence Industry Investment Fund, and the National Integrated Circuit Industry Investment Fund to drive social capital to establish a batch of early-stage venture capital and venture capital funds focused on AI, building a "patient capital" support system covering seed, startup, and growth stages.

At the same time, new support methods such as computing power equity investment, data equity investment, and investment and incubation linkage are being explored.

| Open Source Ecosystem: AtomGit is named with "co-construction and sharing" as the keyword

The document specifically mentions supporting the growth of the national AI open-source community AtomGit, encouraging startups to contribute open-source models, tools, datasets, and other resources.

This signal is clear: the country hopes AI entrepreneurship will not be "fighting alone," but rather forming an open and collaborative ecosystem.

For investors, this means that core contributors and platform-based projects within the open-source ecosystem may receive policy support and resource support in the future.

Kingtech Viewpoint | Inclusive computing power service provider

Computing power is the top essential demand for AI entrepreneurship. The promotion of inclusive computing power policies will directly benefit computing power scheduling platforms, edge computing power facilities, and computing resource pooling service providers. No matter who succeeds in entrepreneurship, they must rely on computing power.

Focus areas: computing power infrastructure operators, computing power scheduling platforms, edge computing enterprises.

Vertical industry data service providers—"data equity investment" opens new valuation space

The document proposes "data equity participation" as a new support method, meaning that high-quality industry datasets themselves can become venture capital. Vertical data service providers focused on fields such as intelligent manufacturing, biomedicine, and materials R&D are expected to upgrade from "selling data" to "investing in startups through data," opening up new business models and valuation spaces.

Focus areas: industrial data platforms, industry dataset construction enterprises, data annotation and governance service providers.

AI incubators and entrepreneurship service platforms—the "investment-incubation linkage" model is worth noting

The document encourages incubation carriers to break away from traditional office space leasing models, incorporating computing power supply, data sharing, and scenario integration into core services, and exploring mechanisms such as "rent equity participation" and "investment-incubation linkage." This means that outstanding AI incubators have the opportunity to evolve from "collecting rent" to "collecting equity," deeply binding with startups.

Focus areas: technology park operators and entrepreneurship service platforms with AI professional incubation capabilities.

The essence of this document is to reduce the supply-side costs of AI entrepreneurship—cheaper computing power, easier data access, easier scenario integration, and more ample early-stage capital. But whether a startup can succeed ultimately comes down to commercialization validation.

A recent research report by CITIC Securities also pointed out that since 2026, the investment logic for the AI industry is shifting from "competition between model capability and capital expenditure" to "commercial verification of orders, revenue, and profitability." Policies can help lower the starting line, but whether you can win depends on whether the product truly solves customer problems.

For investors, rather than betting on whether a single "one-person company" can succeed, it's better to bet on the "infrastructure" along the industrial chain: computing power, data, scenarios, and incubation.