What are the key factors behind the success of China’s DeepSeek program?

DeepSeek is a prime example of innovation and creativity in the technology sector. From a technical perspective, what stands out most about DeepSeek is that it was born out of China’s restrictions and limited access to advanced technologies, including powerful GPUs like the Nvidia A100 or H200, due to sanctions.

Before DeepSeek emerged, the global AI industry seemed to have been “lulled” into believing that AI development required heavy investment in high-end hardware. Leading tech companies like OpenAI, Google, and Meta all pursued this path, driving Nvidia’s stock prices to soar due to the increasing demand for AI hardware. It was widely believed that the future of AI depended on powerful and expensive computing systems.

However, DeepSeek has proven there is another way. Instead of engaging in a race for massive hardware power, DeepSeek focused on optimization and innovation within constraints. These limitations pushed China to find its own development path rather than following other countries’ models.

I believe DeepSeek leveraged several key factors to achieve success.

First, it capitalized on existing research. Rather than developing everything from scratch, DeepSeek strategically utilized previously published studies and innovations. Second, it simplified its model and optimized resources. Unlike OpenAI’s GPT-4, which relies on highly complex algorithms and massive models, DeepSeek adopted a more streamlined approach. Third, one of DeepSeek’s biggest differentiators is its open source approach. This aspect has been greatly welcomed by the global research community.

What does the story of DeepSeek say about China’s AI strategy?

In my view, the story of DeepSeek is the outcome of a well-planned strategy that China has been implementing for years.

China has long invested in education, science, and talent development. The country has been willing to pour significant resources into universities, support top experts, and build a strong knowledge foundation early on. One of its key strategies is the “talent program”, which attracts Chinese-origin experts from around the world to contribute to the country. Thanks to this strategy, China has gradually accumulated expertise and built a strong domestic talent pool. With a solid foundation in place, they can develop their own technology rather than relying on external sources.

DeepSeek has emphasized that nearly 100 per cent of its development team consists of Chinese professionals who were educated, trained, and are working in China, creating products for the domestic market. This highlights the core message of DeepSeek’s story: talent cultivation.

If a country only purchases technology from abroad, no matter how much money is spent, it will only grasp the surface level of the technology without fully understanding its operation and optimization. This leads to dependence on foreign experts and results in massive ongoing costs.

On the other hand, by investing in human capital and training top experts domestically, a country can truly master its technology.

Regarding Vietnam, where does its AI ecosystem stand regionally and globally? What key challenges is it facing?

Vietnam is still at a low level, mainly due to the lack of a clear strategy and long-term vision. Most companies follow global trends rather than developing their own approach.

The priority now is to build a team of true experts who can assess the country’s strengths and weaknesses. Strategies must focus on Vietnam’s unique advantages, with targeted investments. For weaker areas, relying on foreign technology may be necessary, but at the same time the domestic tech ecosystem must be protected and developed.

Businesses also need to invest more in AI. Large corporations already do, but many small and medium-sized enterprises (SMEs) have not. Some previously avoided AI, seeing it as too complex, while others now think hiring an engineer is enough to build advanced models. Vietnam must define its role in the AI process instead of vacillating between these two extremes.

To move forward, more forums and discussions are needed, where the government can guide businesses on effective AI investments. This will be key to building a sustainable tech ecosystem.

As you mentioned, DeepSeek benefits greatly from open source technology. What opportunities does this present for Vietnam? Can Vietnam leverage open source in a similar way to achieve a breakthrough?

Vietnam can learn from China in some aspects but must also recognize its own capabilities. We cannot simply replicate China’s model or build world-class systems to compete with DeepSeek or OpenAI. The key is to identify Vietnam’s strengths compared to China and find niche areas for development - opportunities that major tech firms may overlook because they are focused on bigger challenges.

Vietnam should not aim to create another DeepSeek. AI is a playing field dominated by tech giants. While Vietnam can participate, it requires immense commitment and sacrifice. The real question is: does Vietnam need to follow this path? Should it chase after creating a “second DeepSeek”, or instead focus on resolving its own specific challenges and developing AI systems that cater to its own needs rather than copying OpenAI or DeepSeek?

What should Vietnam do?

First, in the domestic market, Vietnam can develop AI products tailored to its own unique strengths and weaknesses. These solutions can address local challenges that foreign platforms may not be optimized for.

Second, to expand its influence internationally, Vietnam can target markets with similar conditions but less AI development, such as Bangladesh, Sri Lanka, and certain African nations. This approach allows Vietnam to carve out its own space without directly competing with global giants or solely relying on their products. The key is to identify the right target markets.

To achieve this, Vietnam must first secure its domestic market and prioritize sectors that require local expertise. At the same time, it needs to recognize countries facing similar challenges and explore strategic expansion opportunities.

In AI, only the top player truly wins; coming second often means failure. But beyond that race, there are many other opportunities. The goal doesn’t always have to be reaching number one, as that is nearly impossible.

If Vietnam simply follows the path set by others, it will always be dependent and lag behind. Instead of chasing the tech giants, why not focus on different challenges? These are the areas where Vietnam can create real value for itself.

Countries like South Korea and India have implemented strong policies to boost domestic AI development. Should Vietnam develop AI independently or prioritize partnerships with major tech firms? What is the best strategy for its AI growth?

A balanced approach is necessary. The challenge is to move fast, keep costs low, and ensure security. Vietnam must leverage opportunities presented by big tech companies, as they are also looking for customers like the Vietnamese Government to buy their technology. In the short term, Vietnam can take advantage of these solutions.

However, full reliance on foreign technology is not an option. Vietnam needs to gradually build its own AI ecosystem, utilizing local talent and major domestic tech firms such as FPT, Vingroup, and Viettel. These companies should be given incentives to develop core AI systems, especially in critical sectors like education, healthcare, and national security. While collaboration with foreign partners is possible in some areas, there must be clear boundaries between what needs to be safeguarded and what can be open for cooperation.

Additionally, a well-structured talent development and retention strategy is crucial. Vietnam must have a plan to attract and retain top AI talent, ensuring that skilled professionals contribute to the country rather than working for foreign companies or relocating abroad. Without a strong talent strategy, Vietnam risks becoming dependent on foreign AI technology, lacking the ability to develop its own solutions.