A year ago, the explosive arrival of DeepSeek in the artificial intelligence market made a huge impact. This Chinese startup promised to revolutionize AI with cost-effective models, challenging the colossal investments of American tech giants. Many predicted the collapse of the AI infrastructure market and a destructive price war. Yet, twelve months later, the reality is quite different. Not only have investments in AI skyrocketed, but DeepSeek has paradoxically contributed to accelerating innovation and the adoption of new technologies. I invite you to analyze together this fascinating phenomenon that perfectly illustrates the complex dynamics of the modern artificial intelligence market.
📋 Summary
The Paradox of Massive Investments
Contrary to pessimistic predictions, the year 2025 was marked by an explosion of investments in artificial intelligence infrastructure. American tech giants – Amazon, Microsoft, Alphabet, Oracle, and Meta – collectively invested $443 billion, a spectacular increase of 73% compared to the previous year. This phenomenal growth 💰 demonstrates that the DeepSeek effect did not discourage investors; quite the opposite.
This trend can be explained by a well-known economic phenomenon: the Jevons Paradox. When a technology becomes more efficient and less expensive, demand increases proportionally, paradoxically leading to a rise in total consumption. In the case of AI, the reduction in inference costs has paved the way for new use cases that were previously unthinkable.
Companies quickly realized that this demonetization of costs allowed them to explore innovative applications such as real-time video translation, agentic AI, and the automation of complex processes. This diversification of uses has created exponential demand for computing resources, justifying the massive investments observed.
Innovation through Technical Optimization
DeepSeek did not revolutionize AI through a single disruptive innovation, but rather through a smart combination of existing optimization techniques. The Chinese startup masterfully assembled several methods such as model distillation, attention matrix compression, mixture of experts (MoE), and low-precision arithmetic. This holistic approach 🔧 has achieved remarkable performance with limited resources.
This optimization strategy has inspired the entire tech ecosystem. American hyperscalers quickly integrated DeepSeek models into their respective cloud platforms. AWS Bedrock, Azure AI, and Vertex AI now offer DeepSeek-R1 and V3 models, demonstrating the recognition of their technical value by market leaders.
The impact of this approach goes beyond the technical framework. It has shown that it is possible to achieve cutting-edge performance without necessarily having the most advanced hardware resources. This lesson has encouraged many companies to rethink their innovation strategies, favoring smart optimization over raw infrastructure investments.
The Rise of Chinese Open Source
The most significant DeepSeek effect concerns the emergence of China as a major power in the development of open-source AI models. This strategy contrasts with the proprietary approach of American companies and reveals an underlying geopolitical battle for control of the AI ecosystem. Chinese companies like Moonshot AI with Kimi K2.5 or Alibaba with Qwen3-Max-Thinking now claim to have performance 🚀 superior to American benchmark models.
This rise is accompanied by a deliberate openness strategy. By making their models open source, Chinese companies aim to rapidly develop their ecosystem and catch up technologically. This approach is reminiscent of AMD’s strategy against Nvidia: using openness as a competitive lever against an established leader offering closed solutions.
Recent statements by Demis Hassabis, CEO of Google DeepMind, confirm this trend. According to him, Chinese models are now “just a few months” away from American standards. This rapid technological convergence is redefining geopolitical balances in the artificial intelligence sector and forcing Western players to rethink their competitive strategies.
The Challenges of Monetization and Full Stack
Beyond technical performance, the crucial question remains that of monetization and value creation in the AI chain. Companies must determine which links – semiconductors, cloud, models, or applications – will generate the most revenue in the long term. This strategic reflection becomes even more important as competition intensifies 💼 across all segments.
The “full stack” approach seems to be emerging as the winning strategy. Companies that can innovate harmoniously at all levels – from chip to application – have a decisive competitive advantage. Google perfectly illustrates this approach with its own chips, cloud infrastructure, proprietary models, and consumer applications.
This need for vertical integration explains why China is investing heavily in the development of its own specialized semiconductors. Limited access to the most advanced Nvidia chips is pushing Chinese companies to develop their own solutions, as evidenced by the upcoming IPO of Kunlunxin, Baidu’s semiconductor subsidiary.
The Impact on New Use Cases
The drastic reduction in inference costs – by a factor of 20 to 50 compared to GPT-4 – has unleashed considerable innovation potential. Applications that were previously economically unviable have become accessible, creating new markets and opportunities. Agentic AI, real-time video translation, and continuous big data analysis are just a few areas that directly benefit from this 🌟 democratization of costs.
This evolution fundamentally transforms companies’ approaches to artificial intelligence. Rather than limiting themselves to a few premium use cases, they can now consider massive integration of AI into their daily processes. This generalization of use creates a virtuous circle of innovation and continuous improvement.
The impact is also felt in the field of marketing automation. Companies can now deploy sophisticated AI agents to manage their campaigns, analyze performance, and optimize their strategies in real-time. This silent revolution is transforming digital marketing professions and redefining the skills needed to succeed in this sector.
Market Evolution Prospects
An analysis of the DeepSeek effect reveals structural trends that will shape the future of the AI market. The race for efficiency is gradually replacing the race for raw power, favoring innovative optimization approaches. This evolution benefits agile players capable of combining technical creativity and long-term strategic vision 🔮.
The geopolitics of AI is becoming more complex with the emergence of distinct technological blocs. On one side, the American ecosystem relies on vertical integration and proprietary models. On the other, the Chinese approach favors openness and community collaboration to catch up. This duality will likely create two parallel ecosystems with their own standards and practices.
The future likely belongs to companies capable of navigating between these two worlds while developing their own competitive advantages. Mastery of the full stack, from chip to application, becomes a prerequisite for maintaining a position in this ever-evolving market. European companies, in particular, will need to define their own strategy to avoid suffering from this technological bipolarization.
Conclusion
The DeepSeek effect perfectly illustrates the complexity and dynamics of the modern artificial intelligence market. Contrary to initial predictions, this Chinese innovation has not destroyed the market but has paradoxically energized it. I observe that this silent revolution has created new balances, fostered innovation through optimization, and accelerated the adoption of AI across many sectors.
This analysis demonstrates the importance of looking beyond appearances to understand the underlying economic mechanisms. The DeepSeek effect teaches us that innovation does not always come from technological disruption but can arise from the smart combination of existing techniques. This lesson is particularly valuable for companies seeking to leverage artificial intelligence without necessarily having unlimited budgets.
📝 In Brief
- Investments in AI exploded by 73% in 2025 despite the arrival of DeepSeek
- The reduction in inference costs created new use cases and stimulated demand
- China emerges as a major power in open-source AI against American proprietary models
- The full stack approach becomes crucial for maintaining a sustainable competitive advantage
