The world of Artificial Intelligence (AI) has experienced a roller coaster of emotions in recent months. After a period of euphoria, marked by multi-million dollar investments and dazzling promises, a small Chinese company called DeepSeek has burst onto the scene, turning the game board upside down and sowing uncertainty about the future of AI. Are we facing a true revolution or just a passing fad?
Until recently, tech giants like Google, Meta and OpenAI dominated the scene, competing in a frantic race to develop the most powerful AI models. Their strategies were based on massive resource accumulation, investing fortunes in the acquisition of thousands of processing chips, especially the coveted NVIDIA GPUs. The mantra seemed to be: "the bigger, the better." And investors, dazzled by the prospects of a future dominated by AI, were constantly pumping capital into the market, inflating a bubble that seemed to have no end.
But the emergence of DeepSeek after last Christmas has been a real earthquake. Its AI model, DeepSeek-V3, has proven capable of obtaining results comparable to those of its North American competitors, but with one crucial difference: it requires a tenth of the computing power. While the big tech companies needed around 16,000 NVIDIA chips to train their models, DeepSeek has managed to do so with just 2,000. The announcement caused an immediate domino effect: NVIDIA shares plummeted, and the excessive optimism that reigned in the sector turned into nervousness and uncertainty.
The key to DeepSeek's success seems to lie in efficiency. Faced with a shortage of resources and economic constraints, the Chinese company has opted for innovation and optimization. Instead of following the Silicon Valley route, they have explored alternative paths, developing more efficient algorithms and making the most of the power of each chip. It is a palpable demonstration that ingenuity can be a more powerful weapon than simply accumulating resources.
Another fundamental aspect that differentiates DeepSeek from its competitors is its open source philosophy. While large companies keep their models under strict industrial secrecy, DeepSeek has chosen to share its code with the scientific community. This decision, apparently risky, could be the key to its long-term success. By opening its technology to the world, DeepSeek is encouraging collaboration and accelerating the development of new applications. An innovation ecosystem is being created around its model, which could make it the de facto standard in the future.
Although it must be acknowledged that its code is not entirely open, it is what is called “Open Weight”, and can be viewed but not modified. It remains to be seen whether it ends up being completely open as its creators claim.
The impact of DeepSeek's emergence is still difficult to gauge. Are we facing a paradigm shift in the development of AI? Will efficiency and collaboration prevail over opulence and secrecy? It is too early to give a definitive answer. Time, as always, will be the implacable judge.
However, we can draw some preliminary conclusions. DeepSeek's strategy has revealed the vulnerabilities of the prevailing model in Silicon Valley. The dependence on huge amounts of hardware and the lack of transparency could become a burden for large technology companies. On the other hand, the commitment to efficiency and open source opens up a new horizon of possibilities, democratising access to AI and accelerating its development.
The future of AI is exciting and uncertain. The emergence of DeepSeek has injected a dose of realism into a sector that seemed to have become disconnected from the laws of economics. Competition, innovation and collaboration will be the keys to determining who will be the winners in this technological race. And, ultimately, who will be the ones to reap the fruits of a technology with the potential to transform the world.
As I almost always say: time will dictate its final verdict.
EN 
ES