We use cookies to offer a better browsing experience, analyze site traffic, personalize content, and serve targeted advertisements. By clicking accept, you consent to our privacy policy & use of cookies. (Privacy Policy)

DeepSeek: A Valid Threat or Another Overhyped Startup?

Michael Abadha Blockchain market writer
    Summary:
  • DeepSeek, the Chinese startup that was unknown a week ago has taken the world by storm, sending US tech stocks down by over $1 trillion.

The world woke up to news of a major storm in the tech space. Ironically, the trigger came from a fringe company- DeepSeek, a Chinese startup that was largely unknown a week ago.

They say still waters run deep, and that was evident yesterday, as the company was responsible for the wiping out of approximately $1 trillion from the market capitalization of major tech companies like Nvidia, Microsoft, Oracle, Alphabet, Broadcom, Super Micro Computer, Taiwan Semiconductor Manufacturing Company (TSMC) among others.

What is DeepSeek?

While it only came to the fore this week, DeepSeek is a Chinese startup founded in 2023 by Lian Wenfeng. As the US banned exports of AI chips to China, the startup was building R1, its Large Language reasoning model.

While the likes of OpenAI, Meta and Microsoft spent billions of dollars in acquiring high-end AI chips for training their models, DeepSeek used cheaper chips available at their disposal to train its AI assistant.

The end result? the Chinese startup developed a model that needed a small fraction of the computing power used by American tech giants. $6 milllion later, it is able to offer AI chatbot functionalities with the same capabilities as ChatGPT.

How Has DeepSeek Disrupted the Market?

By offering Large Language Model (LLM) training at a fraction of the standard market cost, DeepSeek has sparked a revolution in the tech space. Tech giants had spent billions of dollars as they competed in acquiring high-end AI GPUs. For instance, Meta announced recently that it planned to spend $65 billion toward its AI-related projects. Microsoft has set aside $80 billion towards the same.

Many other tech companies have allocated tens of billions of dollars towards the same. However, DeepSeek’s undercutting approach will likely send these companies back to the drawing board. While a lower capex should typically spell good news to a company’s bottom line, it could still prove costly to reconfigure revenue streams to align with the reality of low-cost AI ecosystem.

DeepSeek’s Limitations

While it has been lauded for being able to go head-to-head with its bigger rivals from the West, DeepSeek is already drawing criticism for its conservative approach to certain topics. For instance, it is unresponsive to queries touching on controversial political events in China like the Tiananmen Square violence of 1989. That is seen as an attempt to limit freedom of speech and control access to information. Some users might consider that a major turnoff for an AI assistant.

Also, some opine that DeepSeek’s capabilities are yet to be tried on a global scale. Therefore its true ability to compete with ChatGPT, Gemini, Llama etc could have been blown out of proportion. That said, DeepSeek passed ChatGPT as the number one downloadedc app on App store on Monday, sending a strong message on its capabilities.