DeepSeek just flipped the script on the AI industry’s narrative that more money and power are the keys to success. Here’s the breakdown, for the past few years, the AI conversation has gone like this:
AI Founder: “Look at this AI that can write a book summary!”
Investors: “This is the next big thing!”
Skeptics: “This is going to mess everything up.”
Tech firms: “We need more funding!”
The public: “Can we avoid wrecking the environment over this?”
Tech firms: “We need more resources! More power!”
But in 2025, DeepSeek comes along and challenges the whole idea that AI projects need tons of money and power. The Chinese company just dropped their model, R1, which is making waves across Silicon Valley and Wall Street. It’s a cheaper, more efficient version of ChatGPT, built with a fraction of the budget and fewer chips than other leading AI models.
DeepSeek built their model for under $6 million, compared to the $100+ million it took for GPT-4, and used just 2,000 Nvidia chips—far fewer than OpenAI’s 25,000. This is shaking things up and making investors rethink their whole strategy.
For Nvidia, this is tough news. The company’s shares dropped 17% on the news, losing $600 billion in market value in a single day—making it the biggest one-day loss for any stock ever. The overall market took a hit too, especially the tech-heavy Nasdaq.
But while one bad day doesn’t signal the end, DeepSeek is making investors rethink the huge investments in AI infrastructure, like building massive data centers and relying on thousands of expensive chips. It turns out, scaling up data centers might not be the most cost-effective way to go.
The AI industry has been following two paths: one focused on better algorithms and reasoning, and another pushing for huge energy and chip investments. According to DeepSeek, the second path is wasteful and unnecessary.
AI isn’t going away, but the landscape is changing quickly—especially for companies like Nvidia, which has been the go-to supplier for AI infrastructure. If DeepSeek’s success with fewer resources proves anything, it’s that AI companies might not need as many chips as previously thought.
But let’s not jump to conclusions. Nvidia is still the big player in launching broader AI infrastructure, so this competition in the AI space is far from over.