Listen to the audio version of this article (generated by AI).
Two months, $10 million, and a team of less than 200 engineers.
That’s all it took for one Chinese company most of us had probably never heard of to trigger an existential panic in the AI race.
Back in January 2025, DeepSeek managed the unthinkable. The startup significantly reduced training expenses for their R1 AI model at a time when spend among top players from OpenAI to Anthropic was already surpassing the GDP of several small countries.
And here’s the kicker: DeepSeek built their models using weaker chips—way weaker.
