The Economics of AI
Yesterday, the release of DeepSeek’s R1, a generative AI platform developed in China, left Wall Street trembling. The Asian tech company not only built an incredibly efficient AI model, able to compete with its leading counterparts, but did it for a fraction of the costs. According to its developers, they spent only US$5.6 million.
The meagre figure is likely an exaggeration –China provides massive non-cash benefits for tech industries. Still, DeepSeek’s main competitors –OpenAi, Google, and Meta– have invested billions in their AI products, and only this year, they have committed US$700 billion in their development. Will it ever pay off?
The challenge is enormous. For comparison, investment in computers only became financially viable in the 1990s, thirty years after they first emerged. By then, according to The Economist, computers had become productive and cheap enough for companies to take the risk and replace human labour. Yet, along the way, billions of dollars were invested, growth was sluggish, and thousands of companies ran out of resources, abandoning the race.
The future of the AI industry could follow a similar path. For now, only 6% of American companies use AI in its operations, and although Nvidia and others have shown improving numbers, they are far behind its capital investment. So far, companies have struggled to imagine practical ways to use this technology effectively.
In this area, DeepSeek may have an upper hand. By providing an incredibly cheap product, it could show green numbers straight away, with companies benefiting from cheaper access to the tool. Likewise, by using simpler chips, it has shown to the world that high-end Nvidia chips are not essential to develop the industry. This is the reason why Nvidia shares tumbled.
Yet, that is far from accomplishing the goal: A product that could productively and cheaply replace millions of minds. In that respect, American companies may lead the way, especially because they have access to more complex chips, which have been banned for Chinese makers. That does not mean, however, that some may not abandon the race along the way.
