Author |Wan Lianshan

Recently, South Korea fired the first shot in the global "AI public service" initiative.

According to the "Universal AI" plan announced by the Korean Ministry of Science and ICT, the government plans to provide free and unlimited generative AI services to its 51 million citizens, clearly promising "no subscription fees and no limits on token usage."

Where will the money come from? Directly integrating with medical appointment systems, public housing application systems, and tax systems, with operating costs covered directly by the fiscal budget.

Progress has already moved from slogans to selecting individuals for work: On August 28, three consortia led by SK Telecom, Kakao, and KT were selected; on September 4, the government convened a project launch meeting to advance the provision of services to the entire population by the end of the year.

On the surface, this seems like the government treating the citizens.

However, upon closer examination, it is far more than just a simple "welfare" initiative.

AI is a tool, and will increasingly become a means of production.

By funding and guiding policies for nationwide AI usage, it amounts to distributing future means of production to every individual.

Regardless of the outcome, this is undoubtedly a grand social experiment.

01 First, let's look at the practical issues.

South Korea has indeed benefited from the hardware dividends of the AI era, but who has heard any news about local large models from South Korea?

According to data from Wiseapp·Retail: In April 2026, there were approximately 23.45 million monthly active users of ChatGPT in South Korea, around 8.45 million for Gemini, and about 2.41 million for Claude.

What about local models? They are virtually ignored.

Moreover, AI competition has a clear scale effect. Without users, products are hard to improve, and revenues cannot support R&D; as products fall behind, users become even less willing to engage.

In a downward spiral, the most skilled ability of startup teams may end up being just modifying funding proposals.

Thus, the "Universal AI" plan stipulates: At least 50% of the services must use South Korean models that meet autonomous foundational model standards, and at least 30% must use models developed by other South Korean companies; foreign models can be used in a limited manner only for necessary functions, but those parts will not receive government support.

At the same time, public procurement, industrial support, and technological sovereignty have been intertwined.

Local companies will gain computational power support and application scenarios, citizens will receive free tools, and the government hopes to retain local R&D, operation, and service integration capabilities.

The government funds the project, locking 80% of computational power for domestic models, effectively helping local industries overcome the most challenging initial stage.

How much is the funding?

On September 4, MoneyToday reported based on information from parliamentary offices that the planning for 2027-2030 is 250 billion won annually.

170 billion won will be used for GPU leasing, 30 billion won for the three consortia, and 50 billion won for expanding the AI ecosystem.

With South Korea's population around 51.61 million in 2026, this amounts to about 4,800 won per person per year.

This amount is clearly insufficient, even if doubled it still wouldn't be enough.

So why does this budget work?

It relies on a mix of different usage intensities and the allocation of computational resources for different tasks.

SK's proposal published on August 19 specifically stated that simple problems should be handled by lightweight models, while complex tasks should be addressed by high-performance models.

Model routing combined with caching, batch processing, and quantization optimizations can reduce the cost per task.

Services are also preparing to retain commercial revenue.

According to a report from Financial News on September 7, the government is considering allowing moderate advertising and discussing charging for premium services after the second half of 2027.

Therefore, the fiscal path should be: initially investing in starting computational power, subsequently supporting services with a planned 250 billion won annually, and relying on a broader range of AI R&D and infrastructure investments.

Of course, the final total cost will include enterprise contributions and capacity expansions resulting from actual usage, which cannot be estimated at present.

However, once this path is established, South Korean enterprises will inevitably gain an advantage in domestic governance and life services.

To turn this convenience into economic growth, South Korea still faces a challenge.

02 According to research data published by the Bank of Korea in June 2026, the average use of AI shortens working hours by 3.8%, equivalent to about 1.5 hours per week, corresponding to about a 1.0% potential improvement in productivity.

However, the study found no evidence that saved time universally translates into real output growth.

The effects are notably prominent among self-employed individuals, professionals, and high-intensity AI users.

This is the productivity J-curve seen in the diffusion of general technology: tools first enter, followed by training, process reorganization, and organizational adjustments, before benefits gradually materialize.

Universal free access can lower the technical threshold, but to truly unleash output, businesses must be willing to adjust work methods, and employees must be motivated to save time.

Otherwise, while the average "AI degree" increases, average overtime may just as likely rise.

This is what we often discuss recently: AI is becoming stronger, but people are becoming more exhausted.

Deeper changes will occur in the distribution of benefits.

The government pays for users, and the first to gain certainty in demand are computational power suppliers, model companies, and platform operators.

For small businesses, public platforms may reduce development and customer acquisition costs; for consumers, they may lower transaction costs.

However, when chat interfaces further control appointments, shopping, and service recommendations, platforms will also gain new distribution rights.

Who gets recommended, who pays commissions, and who can access will all impact the competitive landscape.

This is also why fiscal subsidies must consider usage outcomes.

According to the report by Financial News on September 7, GPU support will be allocated based on user and usage performance, with the first evaluation considering March 2027.

In practical execution, task completion rates, error rates, and actual time saved must also be included in evaluations.

The labor market will adjust accordingly.

AI assistance may help inexperienced individuals complete more tasks, but it may also compress jobs originally available for newcomers to practice.

Older individuals may have easier access to services, while younger individuals’ ability to continue accumulating skills and earning income will similarly determine how far this reform can go.

After popularizing tools, further training, job adjustments, and income distribution must follow.

The current judgment is that South Korea is in a position to make universal AI a practical public service, thereby expanding the market for domestic models.

Established user channels, financial support, and local service connections give it a foundation to launch.

Other countries may also see a possibility:through public procurement, transforming AI from individual subscription products into universally available social services.

However, enabling the widespread appearance of productivity dividends necessitates a longer organizational and institutional transformation.

This is a more profound change.

03 The greatest aspect of this experiment is the attempt to provide the most crucial production tools of the future to every citizen as a public service.

The Ministry of Science and ICT of South Korea clarified in the project announcement that the continuous upgrade direction states "one AI agent for each citizen," allowing citizens to participate in economic and social activities through their own AI agents and share the benefits brought by AI.

Distributing cash increases current purchasing power; providing production tools aims to enhance the ability to generate income in the future.

The role of fiscal expenditure has thus extended from subsidizing consumption to supporting ordinary people’s participation in production.

For example, a young person wanting to run a small business needs to promote, organize customer demands, analyze orders, and occasionally make modifications to the website.

Even if they can use AI tools, the tokens consumed can still cost quite a bit.

If public AI can assist in completing part of this workload, their minimum investment needed to try starting a business could decrease.

This not only lowers entry barriers but also expands the production resources one can mobilize.

A person’s experiences, judgments, and customer relationships, combined with accessible and effective tools, could support businesses that would otherwise not have been feasible.

This change may also affect the bargaining power of workers.

If an individual can only rely on company-provided software, equipment, and organizational systems to work, their costs of leaving the job will be relatively high.

If they can leverage public AI, they could independently take on some work or more easily learn new skills and switch careers, thus gaining more external options when negotiating with employers.

Having the ability to choose a different path will make them firmer when discussing wages.

This also pushes forward the issue of social distribution significantly.

When discussing income distribution, we often think about how wealth is adjusted through taxation and welfare after it is created.

Universal AI, however, aims to intervene earlier:to help more people obtain the tools to participate in creating wealth before wealth is generated.

Furthermore, the question of how ordinary people engage in growth arises:whether as consumers purchasing products, or as individuals capable of using new tools, providing services, and running businesses.

By covering part of the tool costs, the government lowers the restrictions on obtaining productive capacity based on family background, company size, and individual income.

Once ordinary individuals possess a production tool, they gain a new livelihood path and more confidence in seeking employment, ensuring they benefit from the gains of technological advances.

Everything will be better.