👉 Background: DeepSeek, founded in 2023, released their first large language model later that year. Earlier this year, DeepSeek released its latest model, called "DeepSeek R1", which claimed to be on-par with ChatGPT. Shortly after its release, DeepSeek’s AI assistant became the No. 1 downloaded app in the US App Store.

‍👉 What happened: DeepSeek's researchers claim that it cost only $5.6 million USD to train their foundational model, using just over 2,000 Nvidia AI chips. Compare that to Meta, which has currently used 600,000 Nvidia AI chips. This rattled the whole of Wall Street, particularly Nvidia investors who caused a 17% plummet in its share price.

‍👉 What else: DeepSeek has managed to produce such good results at a fraction of the cost of its US competitors. As a result, the spotlight was firmly placed on US companies who have invested billions of dollars in resources and computer chips.

What's the key learning?

💡When a disruptive technology emerges, it forces both suppliers and competitors to reassess their business models.

💡Analysts have expected Nvidia, the AI chip designer, to increase their AI chip shipments by 8-10% in 2025. But all of a sudden, DeepSeek has been able to do more with fewer chips, and create a similar model for just over $5 million USD. That means demand for Nvidia’s AI chips could shrink significantly.

💡On the other hand, DeepSeek's competitors, like OpenAI, have invested billions into developing advanced AI models - but their capabilities have been matched for a fraction of the cost. And this means DeepSeek can offer consumers a range of AI products, that are comparable to OpenAI, for much less. All in all, the AI market is the US has received a big wake-up call. Your move, OpenAI.