DeepSeek reportedly completed a model training run for roughly $6 million and released its technical work, raising a question that deserves serious discussion: why haven’t governments made national open models a long-term investment priority?
What this is
An “open model” is an AI model whose weights, or some of its training methods, users can download, deploy, and inspect. Compared with commercial models that only expose an interface, open models give companies, research institutions, and individuals greater control over how the system runs. The original Reddit poster argued that the United States has not trained a national open model, while DeepSeek has already released one. But training cost is only one line item. It excludes data governance, compute procurement, safety evaluation, ongoing maintenance, and infrastructure investment. Claims of a “very low cost” therefore make more sense as a comparison of a single training stage than as the full cost of developing and sustaining a model.
Industry view
The case for public investment goes beyond technological autonomy. Open models can reduce companies’ dependence on a small number of compute platforms and APIs, limit the risk of sensitive data crossing borders, and provide modifiable foundations for scientific research, education, and small and medium-sized businesses. DeepSeek’s emergence further shows that the relationship between model capability and training expenditure is not necessarily linear.
The risks are equally clear. National programs can turn technological competition into administrative procurement, and open does not automatically mean secure or reliable. Released weights could also be used to build fraud tools. More importantly, subsidizing large models alone may not produce a complete ecosystem spanning compute, talent, data, and applications. The real question is whether government funding can sustain long-term open collaboration rather than simply purchase a “domestic large model” for display purposes.
Impact on regular people
For enterprise IT: More deployable open models give companies a way to compare the long-term costs of cloud APIs with local deployment, while reducing the impact of outages involving a single supplier.
For individual careers: Whether new jobs appear will depend on whether models can be integrated reliably into business operations, not on whether the government releases a model. Data compliance, auditing, and accountability for errors will move much faster onto procurement checklists.
For the consumer market: Open models may lower prices for assistants and software services, but security incidents, copyright disputes, and differences in service quality mean low-cost products will not necessarily become the default choice.