Chinese startup DeepSeek challenges AI spending norms
DeepSeek’s claims challenge the narrative that scaling AI requires vast investments in computing power.
Chinese startup DeepSeek has claimed to build an advanced AI model at a fraction of the cost and time required by US tech giants, raising questions about the industry’s spending. The company stated it developed its V3 AI model in just two months for under $6 million, using Nvidia’s less-advanced H800 chips. An app powered by this model became the most downloaded app on the US iPhone store on Monday.
Founded in 2023, DeepSeek claims its AI models rival or outperform top competitors despite significantly lower costs, challenging the notion that scaling AI demands massive resources. This comes as US tech firms, part of the “Magnificent Seven,” have collectively added $10 trillion in market value since the AI boom sparked by ChatGPT in late 2022.
Industry analysts, however, remain sceptical of DeepSeek’s claims. Bernstein‘s Stacy Rasgon questioned whether the startup’s advancements could outpace those from established global AI labs, suggesting the company’s narrative might be overstated. The developments have drawn attention as investors scrutinise AI spending ahead of earnings reports from key industry players.