The emergence of Chinese AI company DeepSeek has been a wakeup call for investors and the tech industry. Legendary venture capitalist Marc Andreessen called it China’s “Sputnik moment.” While the AI race is clearly on, it’s too early to make assumptions about who will ultimately win and the ripple effects the rise of inexpensive AI may have.
It does look like China leapfrogged Western AI technology similar to how Russia launched the first satellite at a time when conventional wisdom said America held a dominant position in the space race. One big question that remains unanswered is how this happened secretly in our modern era of globalization. When the Russians built Sputnik, they were able to do it in isolation. Hardly anything is done in isolation now, so the fact that DeepSeek’s breakthrough was developed under the radar speaks to a great effort to conceal it.
The DeepSeek Advantage
DeepSeek’s technology should not be underestimated. They succeeded in training their AI for roughly 3% of what OpenAI spent to create ChatGPT, and they did it without the top-of-the-line NVIDIA chips, due to U.S. government restrictions on exporting them to China. The fact that they didn’t need those chips is why NVIDIA shed $589 billion market capitalization on the news. It shows that AI doesn’t have to be this incredibly resource intensive thing anymore.
But that’s not the whole story, since the export controls haven’t been in place long enough to really make a difference. They could still hold back Chinese AI development in the longer term. After all, just because a comparable model can be developed for less money, doesn’t mean that greater computing power won’t yield higher performance or further breakthroughs.
Now that DeepSeek has proved that AI doesn’t have to cost billions of dollars—and it has open-sourced its R1 model—the prevalence of AI is likely to increase dramatically, especially with mobile devices. This undermines the competitive moat that companies like OpenAI were hoping to create by building up enormous amounts of energy and computing power. Inexpensive models like DeepSeek’s could be a good thing for the growth of AI generally, even if it calls into question some of the economics of current players. As a result, we can expect to see AI appearing in more places, faster, in the coming years.
Energy Matters
Until now, the convention wisdom was that AI was going to need ever greater sources of energy, including potentially nuclear reactors. DeepSeek’s more efficient technology could undermine this, but it may be too soon to tell, and in fact, the opposite may turn out to be true.
The Jevons paradox is an economic theory that says that greater efficiency actually increases energy demand, rather reducing it as many would assume. The reason for this is that more efficient technologies become more prevalent, faster, requiring more power as a result.
Overall, investors should resist the impulse to jump to conclusions. Russia may have launched the first satellite, but they lost the space race. Competition, like scarcity, often leads to greater creativity, focus, and progress.
The economics of AI may have changed, but that doesn’t mean that the story is over. Right now, AI is in its infancy, similar to the growth of dot-coms during the 1990s. It’s nowhere near mature. And there are also unpredictable political forces at work. While DeepSeek’s technology certainly seems to be the real deal, the timing of the announcement was not a mistake. It came right after Donald Trump’s inauguration as President and followed an announcement of a new mega-fund for American AI development. DeepSeek is a shot across the bow during a brewing trade war.
Rather than the end of AI race, we need to see the emergence of DeepSeek as the beginning of a marathon. This technology will likely evolve many times over, and it’s too soon to call a winner.