Nvidia, a US chip-maker, has expressed its view that cryptocurrencies do not “bring anything useful for society,” despite the fact that the company’s powerful processors have been selling in large quantities in the crypto sector. Michael Kagan, its Chief Technology Officer, has maintained that the uses of processing power such as the artificial intelligence chatbot ChatGPT are more beneficial than mining cryptocurrency. Nvidia has not welcomed crypto miners with open arms; in 2021, it released software that limited the capability of its graphics cards to be used for mining Ethereum, ensuring that its preferred customers, such as AI researchers and gamers, had access to the cards instead.
Kagan asserted that the decision to move away from cryptocurrency mining was justified because of its limited value. ChatGPT version one was trained on a supercomputer comprising roughly 10,000 Nvidia graphics cards. “The crypto mining only required parallel processing, which Nvidia is well known for,” Kagan explained. “They bought a lot of stuff, but eventually it was not seen as beneficial to society. With ChatGPT, everyone can now create their own machine and program it to do whatever they desire. If the results are not satisfactory, you can simply tell it ‘I want something different.’”
Prior to its acquisition by Nvidia, Mellanox, a company founded by Kagan, had seen much success in the high-frequency trading industry. Kagan commented, “We were heavily involved in trading; people on Wall Street were buying our products to save a few nanoseconds on the wire.
Banks were even taking extreme measures, like tightening fibers under the Hudson River to shorten the distance between their data center and the stock exchange.” However, Kagan did not believe that cryptocurrency was something that would do humanity any good. “People do crazy things,” he quipped, “but they still buy your stuff. But don’t redirect the company to support whatever it is.”
Originally renowned for delivering powerful graphics cards to empower PC gamers with cutting-edge game capabilities, it was by chance that Nvidia’s offerings took a pivotal role in the AI revolution. The massively computationally intensive process of training a new AI system, which can cost billions of dollars in computing power, was found to be significantly more efficient when done with the relatively simple, yet powerful processors that had been adopted by gamers.
At its annual conference last week, Jensen Huang, Nvidia’s chief executive, declared the company as the driving force behind “the iPhone moment of AI”. He also expressed that this ‘generative AI’ would have a substantial impact in nearly every industry. Two weeks prior, Microsoft had purchased tens of thousands of Nvidia’s AI-focused A100 GPUs from OpenAI.
Additionally, Amazon had gained 20,000 of the H100 successor chips for its cloud computing AWS service, and another 16,000 have been purchased by Oracle. Nvidia also provides rental access to the chips through its DGX cloud service, which starts at a rate of just under $37,000 (£30,250) a month for eight H100s connected in a “cluster”. Unfortunately, Nvidia’s $40bn takeover of the UK-based tech firm Arm fell through due to regulatory issues.