Deepseek threatens US dominance in AI race and loss giant NVIDIA

Deepseek threatens US dominance in AI race and loss giant NVIDIA

DeepSeek, a rising star in the AI landscape, released extremely efficient AI models that can compete with great products from US companies. A Chinese artificial intelligence (AI) company, DeepSeek, founded in May 2023 by Liang Wenfeng in Hangzhou, China, and backed by the hedge fund High-Flyer, has quickly established itself in the world of AI.

Unlike broad-spectrum models like GPT-4o and o1 which aim to converse fluently on almost any topic. DeepSeek is designed to tackle complex, industry-specific challenges, targeting niche industries such as healthcare, finance, and legal services where accuracy and domain expertise are non-negotiable.

But what are the reasons that have made the US president say that it is a “wake-up call” for US tech firms and NVIDIA’s shares to drop?

DeepSeek has made its ai chatbot open source, meaning its code is freely available for use, modification, and viewing. Chinese artificial intelligence company DeepSeek’s latest large-language model came close to the performance of its American rivals at a fraction of the cost. For instance, DeepSeek reportedly trained its R1 model for approximately $6 million, compared to the $100 million spent by OpenAI on GPT-4.

Deepseek uses a chain of thoughts to give an answer, while most AI models give an answer but not the thought process behind it. It helps in a way that if the Ai model gives a wrong answer, you can track it through reasoning steps to check where it went wrong. Deepseek provides this reasoning model for free while open AI provides this feature for the paid user. But from January 31, free plan users can also try OpenAI o3-mini by selecting ‘Reason’ in the message composer or by regenerating a response. This marks the first time a reasoning model has been made available to free users on ChatGPT.

Chain of thoughts

One of the most remarkable aspects of deepseek is self-evolution. Deepseek in its R1 paper describes the model using behaviors such as reflection where the model revisits and reevaluates its previous steps and the exploration of alternative approaches to problem-solving arise spontaneously. These behaviors are not explicitly programmed but instead emerge as a result of the model’s interaction with the reinforcement learning environment.

That’s why DeepSeek R1 improves over time while OpenAI’s 01 model stays static. That’s how reinforcement learning works. The model explores different ways to answer a question and picks the one that works best.

On 10 January 2025, DeepSeek released its first free chat app, based on the DeepSeek-R1 model, for iOS and Andriod and by 27 January, DeepSeek-R1 had surpassed ChatGPT as the most-downloaded free app on app store in the United States, causing Nvidia’s share price to drop by 18%.

NVIDIA, a technology company that makes chips, systems and software that power things like artificial intelligence (AI), gaming and graphics, is a leading provider of AI computing hardware and has become the biggest company in the world in November 2024. Though for companies like NVIDIA, China is a huge market. Biden administration, limits AI chips export to countries that are close allies of the US while keeping the advanced technology out of the hands of countries like China and Russia. In response, NVIDIA strongly criticized the export controls. They said that it “threatens to derail innovation and economic growth worldwide”.

Moreover, Vice President JD Vance said at an AI summit in Europe on Monday that US will safeguard its AI and chips, and block efforts to “weaponize” them. Trump also called deepseek a “wake-up call for our industries that we need to be laser-focused on competing to win.”

US vice president

What could be the future of this race of going ahead in AI technology where even the U.S. government’s effort to hinder AI development in China through control of chips export, including restrictions on the most advanced offerings from NVIDIA, didn’t stop DeepSeek from creating its generative AI app efficiently, and that also at a level offer by the companies such as OpenAI.?