Jevons Paradox: DeepSeek-R1 Will Ultimately Drive Demand for NVIDIA's GPUs
In 1865, William Jevons observed that coal consumption in England was growing exponentially despite increasingly efficient steam engines. Similarly, in the coming years, more efficient AI models like DeepSeek-R1 could drive demand for NVIDIA's GPUs, not reduce it.
This is the author’s opinion only, not financial advice, and is intended for entertainment purposes only. The author holds a beneficial long position in NVIDIA Corporation. The author receives no compensation for writing this article and has no business relationship with any of the companies mentioned.
In his 1865 work "The Coal Question", the English economist William Stanley Jevons formulated the paradox named after him: "It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption. The very contrary is the truth."
Jevons had studied English coal consumption and found that it had continued to grow exponentially, even though coal had been used more and more efficiently since the invention of Watt's steam engine. Each new generation of steam engines consumed less and less coal than its predecessors, but the number of steam engines in use grew exponentially because the number of applications grew exponentially. At the beginning of the 19th century, steam engines were mainly used to power weaving machines, but over the years the technology spread to almost every other field, disrupting more and more sectors of the economy. Eventually, steam engines were essential to everything from transportation to agriculture, and coal consumption skyrocketed.
In my opinion, NVIDIA's GPUs are today's equivalent of coal in Jevons Paradox, a limited resource. The release of the open source reasoning model DeepSeek-R1 by the Chinese startup DeepSeek led to panicking stock markets and a notable crash in tech stocks on January 27, 2025, led by NVIDIA. The order of magnitude more efficient use of the "resource" GPU made Wall Street question the billions of dollars invested in GPU servers. Until now, it has been assumed that the computing power of massive GPU servers alone created a moat for the development of cutting-edge AI - an assumption that DeepSeek swept away overnight.
Anyone who now sees an investment bubble in GPU server capacity that will be burst by the efficiency of DeepSeek-R1 is, in my opinion, misreading the situation. But how did we get here?
It seems that the Biden administration's export ban was not strict enough to really cut off China from NVIDIA's high-end GPUs, but it was effective enough to make them scarce in China. These conditions may have been the perfect breeding ground for the development of a highly efficient AI model like DeepSeek-R1. While the very existence of this extra-American AI model challenges American technological supremacy, we must not forget that it was trained on NVIDIA's GPUs. As software becomes commoditized, open source DeepSeek-R1 shows me once again that AI is a hardware revolution, not a software revolution. And NVIDIA remains at the center of that revolution.
It may have cost more than the reported $5.6 million to train DeepSeek-R1, and perhaps tens of thousands of NVIDIA's GPUs found their way into China despite export controls. However, it may have been the scarcity of computing power that ultimately drove Chinese innovation to its peak. Necessity was made a virtue: without export restrictions, Chinese AI models might not be much further ahead, but the U.S. would be richer.
Just as Jevons described the consumption of coal by increasingly efficient steam engines over a century and a half ago, the increasing efficiency of AI models will drive demand for GPUs, not reduce it. Efficient open source models like DeepSeek-R1 will lead to a democratization of AI, and AI models will finally become a commodity. More startups and individuals will be able to build AI products themselves. Edge computing will open up new markets that NVIDIA has already entered with products like Jetson. Instead of a few hyperscalers as customers, NVIDIA will sell to more customers and we will see a redistribution of demand. AI will revolutionize all industries, just like the steam engine revolutionized all industries in the 19th century.
The release of DeepSeek-R1 is now often referred to as America's new Sputnik moment. It should be remembered that it was only after the Sputnik shock that U.S. investment and effort in space really took off. So now the AI arms race begins. Nvidia will provide the hardware.
