A plump blue whale, the logo of DeepSeek, emerged under the world’s watch on January 20, 2025, and made waves.

The generative AI chatbot launched by the China-based AI company DeepSeek delivered performance comparable to established AI tools, such as ChatGPT, in a number of important tasks, at a training cost of a small fraction of the latter, using older processors instead of the latest models developed by NVidia.

The impact of DeepSeek was immediately visible. Within a week, DeepSeek surpassed ChatGPT as the most-downloaded free app on the iOS App Store in the United States. By early February, the app became the most downloaded app in 140 markets worldwide.

Another immediate response came from the financial market. The stock of Nvidia, the most valuable company in the world thanks to its leadership in AI processors, plunged approximately 17%, losing almost $600 billion and recording the largest loss of market capitalization in the history of the US stock market.

The selloff was probably a short-term overreaction, In the longer term, innovations in AI and wide adoption of AI could benefit different players in the industry ecosystem, including Nvidia.

However, the rise of DeepSeek does demonstrate the complex, interconnected, and dynamic nature of the emerging global AI industry. DeepSeek, as well as other AI startups worldwide, and more fundamentally, their different underlying approaches to AI, will shape the future of AI.

Disruptive innovation

“Disruptive innovation” is a term first introduced by Professor Clayton M. Christensen in 1995. It refers to an innovation initially ignored or considered inferior by most existing companies and customers; however, by offering a more affordable and oftentimes simpler option targeting an underserved segment, the innovation improves rapidly, eventually outcompeting established players and even changing the entire industry landscape.

Today’s global economy has been fundamentally reshaped by such disruptors, from Apple to Amazon to Alibaba. The rise of DeepSeek, also seems almost a textbook case of the definition of “disruptive innovation”. Several features made DeepSeek’s AI tool uniquely impactful.

Democratization and adoption of AI

DeepSeek’s generative AI model is trained at a significantly lower cost, reportedly at US$ 6 million, compared to the costs of top players, such as OpenAI’s $100 million for OpenAI’s GPT-4 in 2023. The performance of DeepSeek, however, could match or even surpass those of incumbents, including OpenAI’s ChatGPT, not in all tasks but in a number of important specialized tasks.

The performance combined with the affordability has a profound impact on AI adoption. If DeepSeek or other similar AI tools can offer sufficient AI capability at a more accessible price, organizations and nations, including both developed nations and developing nations, that have fewer resources and budgets could access and utilize AI.

Energy and environmental impact of AI

Generative AI is extremely computation-intensive, leading to massive environmental impact. DeepSeek’s model reportedly consumes significantly less energy. Whether it is DeepSeek or future AI tools, in general, if innovations in generative AI can focus on higher efficiency in computing, AI’s environmental impact can be more effectively mitigated.

Policies and politics of AI

AI has become deeply intertwined with politics, policies, and power. The use of DeepSeek has been banned in some institutions or regions, for varied reasons.

Italy became the first country in the world to ban the DeepSeek AI, doing it as early as last month. The app has been removed from app stores in the European country. The ban was imposed by Italy’s privacy watchdog, the Italian Data Protection Authority (DPA), citing concerns on user data.

In the US lawmakers plan to introduce a bill Thursday that would ban DeepSeek’s chatbot application from government-owned devices, citing security concerns regarding user information.

With AI becoming critical to different regions’ politics, the near future of the field of AI will only become even more contested and complex.

Open-source paradigm of AI

Top scientist in the field of AI, however, see the development of AI from another perspective. DeepSeek uses an open-source model, making its algorithms, models, and training details public and allowing its code to be freely available for viewing, use, modification, designing, and further development. Any organization can utilize the code and create their own versions.

Yann LeCun is the Chief AI Scientist at Meta, professor at New York University, and a laureate of the Turing Award, the highest honor in computer science, for his work on AI together with Geoffrey Hinton and Yoshua Bengio. LeCun commented on the impact of DeepSeek:

“To people who see the performance of DeepSeek and think: ‘China is surpassing the US in AI.’ You are reading this wrong. The correct reading is: ‘Open source models are surpassing proprietary ones.‘”

AI for humanity

The emergence of DeepSeek has shaped the field of AI, making incumbents rethink their AI strategies. As technologies, economies, societies, politics, and policies intersect in the field of AI, the future of AI is unknown and uncertain. Let us make sure that despite the complexities surrounding AI, collectively, AI will benefit science, humanity, our planet, and beyond.

***
Alley Oop’s newsletter
Alley Oop arrives in your inbox every Friday morning with news and stories. To sign up, click here.
If you want to write to or contact Alley Oop’s editorial team, email us at alleyoop@ilsole24ore.com.