On the eve of the AI summit in Paris, French President Emmanuel Macron announced an investment of €109 billion as the foundation for France to become a leader in AI, and expects the move to serve as a catalyst for Europe to catch up with Chinese and American AI technology.
However, a few weeks ago, the Chinese AI DeepSeek revealed a technical path for low-cost, high-performance AI. Has Europe drawn inspiration from this, or does it still insist on emulating the highly capital-intensive Silicon Valley model?
For the United States, DeepSeek has sounded alarms on multiple levels. It has almost negated Silicon Valley’s monopoly on innovative technology through a “capital tsunami,” and simultaneously, the orientation of AI values has become an urgent issue in global AI governance.
Regarding the former, it concerns the binding of new technology with business models. Does the world still need to follow the American model—investing enormous funds, establishing technological barriers, and ultimately forming an oligopoly?
Regarding the latter, it concerns Western concerns about China’s technological development—if China becomes the engine of innovation technology, using its powerful productivity as leverage to form a supply chain ecosystem, and then using AI to export values that differ significantly from the West.
These represent two levels of issues: democratization in terms of technological development and democracy in terms of Western values.
The emergence of DeepSeek seems to force the world to choose between two paths. If technological autonomy is to be achieved, one must choose DeepSeek as the platform for developing application products and follow its low-cost, high-efficiency route in large language models, thereby escaping the Silicon Valley model and the oligopoly of American tech companies in the market.
If democracy in Western values is to be upheld, countries or regional alliances outside of China and the United States must choose the traditional Silicon Valley model and then use subsidies to increase capital investment in order to compete with U.S. technology, to achieve as much autonomy as possible, and to avoid being “invaded” by Chinese values.
As the epicenter of the “DeepSeek Earthquake,” the initial panic in the US was reflected in the stock market, and then some of the public opinion started the defense of democracy, claiming that China would expand its values through DeepSeek, which is like a Confucius Institute in the field of AI. Others emphasized the unprecedented challenge to national security.
Regardless, this involves global AI governance—how to manage AI capabilities, prevent this technology from getting out of control, or being abused for political ambitions. This requires international cooperation, especially between the US and China, to draw red lines for AI based on consensus.
Here’s another case where competition and cooperation are both necessary. When the issue involves China, full confrontation is unrealistic. Thus, suggestions have quickly emerged to emulate the cooperation between the US and the Soviet Union during the Cold War on nuclear regulations, setting up safety barriers to allow for fierce competition within a generally safe framework.
However, the nuclear issue is much simpler than AI industry issues because nuclear weapons are not commercial. The aim for the US and USSR was “not to use,” but for AI, the issue is “must use, and how to use.”
India’s reaction is second only to the US. Indians believe they possess the world’s best mathematical talent, so why haven’t they produced a homegrown ChatGPT or DeepSeek? Secondly, like many countries, India welcomes the technological democratization demonstrated by DeepSeek because it means that with less government investment, efficient AI can still be created.
In terms of capital markets, the US leads the world; in terms of government subsidies, only China can compete with the US. Thus, other countries, if they do not want to be technologically monopolized by the US and China, naturally support the DeepSeek approach, opposing technological colonialism and capital bubbles.
Indian entrepreneurs have taken swift action. AIonOS, the AI startup by former Tech Mahindra CEO C.P. Gurnani, has signed an agreement with Indonesian telecom giant Indosat to utilize DeepSeek’s technology to serve the Indonesian market.
Southeast Asian countries have already invested significant government resources and brought in foreign capital to establish numerous data centers, showing ambition in the AI industry. These countries, similar to India, see the true significance of DeepSeek in the democratization of technology, which helps small and medium-sized companies join in and expand the industry scale, and they are not worried about China exporting values through AI platforms, as most application product development on the DeepSeek platform does not involve values.
The key is that the issue of AI exporting values relies on international cooperation to establish norms, which requires long-term, multi-round negotiations. Industry and technology development have their own pace and cannot wait for legal frameworks.
Clearly, Asian countries, especially those with a sense of autonomy, have confirmed that DeepSeek brings opportunities, not poison, and have quickly taken action to seize business opportunities. So, how does Europe, which stands at many crossroads, view DeepSeek?
Simply put, if Europe’s view of DeepSeek is constrained by ideology, it might fall behind India and Southeast Asia in the AI field.
DeepSeek should not be regarded as a “Chinese threat.” In fact, it challenges China’s industrial policies. A unicorn company has achieved the technological goals of domestic tech giants ahead of schedule, despite the latter enjoying government subsidies, which also sounds the alarm for Beijing—is China, like the US, over-investing in AI technology?
Although there are rumors that DeepSeek has received government subsidies, there is no solid evidence, and given DeepSeek’s substantial funding, it actually doesn’t need government subsidies. This fact further motivates other countries and companies, as the essence of democratization is “from the bottom up.” As long as the industry entry threshold is low enough, small and medium-sized enterprises can stand out with technological advantages.
On the other hand, DeepSeek effectively suppresses AI industry bubbles, allowing countries to demand better outcomes with existing investment amounts—even if they pale in comparison to those of the US and China—to avoid over-investment.
Therefore, Europe should, like India and Southeast Asian countries, see DeepSeek as an opportunity rather than a “Chinese threat” to avoid falling behind the East.
If Europe seeks to achieve technological autonomy, it must first reduce its reliance on American technology, which is not easy, but action must be taken. After all, autonomy means freeing oneself from the influence and control of other countries in various aspects, minimizing technological monopolies by foreign entities.
DeepSeek hasn’t completely detached from American technology; it uses Nvidia’s chips. However, DeepSeek has found a workaround within the “Nvidia ecosystem,” achieving its goals with fewer GPUs. This means other GPU manufacturers—including, but not limited to, American ones—can attract large model companies with more efficient design platforms to compete with Nvidia.
In other words, the core technology DeepSeek possesses is a roadmap that allows for switching GPUs at any time, ensuring diverse technological choices. It is expected that China will achieve technology equivalent to Nvidia’s H800 this year, and then DeepSeek will naturally collaborate with domestic GPU manufacturers to prevent interference from US-China tech wars.
Regardless of which country’s technology or company, in a multipolar world, they should avoid being monopolized by any single nation, which should be a global consensus, and DeepSeek unexpectedly provides a pathway.
There’s a playful saying: the US innovates, China copies, and Europe regulates. DeepSeek undoubtedly falls into the category of innovation, encouraging Chinese companies to move away from the “copy” status. How can Europe be content with just “regulating”?
Having said that, “regulation” is also Europe’s leverage. The AI field is still full of unknowns, where Europe can play a balancing role in the governance of AI values, highlighting its importance.
But if Europe cannot see DeepSeek as a revolution for technological democratization or against technological colonialism but is instead constrained by values to view it as a “Chinese threat,” leading to contradictions between policy and industrial interests, it will pay the high price of the Silicon Valley model and lag behind the East.
The answer is still “cooperation.” Technology requires cooperation, drawing red lines requires cooperation, and eliminating ideological barriers and biases as much as possible ensures that technological development does not lag behind.
