Miners who have been in a difficult situation due to a virtual currency crash provide surplus GPUs as ``resources for AI training''
Taking advantage of the virtual currency boom from the latter half of the 2010s, businesses that lend large amounts
Crypto Miners Seek a New Life in AI Boom After an Implosion in Mining - WSJ
https://www.wsj.com/articles/crypto-miners-seek-a-new-life-in-ai-boom-after-an-implosion-in-mining-92a181fd
Crypto Miner Hive Blockchain Touts Privacy of AI Models Running on Its GPUs
https://www.coindesk.com/business/2023/06/30/bitcoin-miner-hive-blockchain-touts-privacy-of-ai-models-running-on-its-gpus/
Miner Pivots 38,000 GPUs From Crypto to AI | Tom's Hardware
https://www.tomshardware.com/news/hive-blockchain-pivoting-to-ai
Many cryptocurrency miners have been hit bythe cryptocurrency crash that occurred from the end of 2021 to 2022, and many of them withdrew due to the merging of Ethereum.
Virtual currency mining using GPU reports that ``no one is profitable'' and minors start to withdraw one after another, adverse effects of Ethereum merging - GIGAZINE
Satoshi Spain , a virtual currency mining company based in Spain, is also one of the mining companies that has grown by selling and leasing mining rigs equipped with powerful GPUs during the virtual currency boom. Satoshi Spain, who was in a difficult situation, started a business that remodeled mining rigs into AI training resources, which has been growing rapidly in recent years, and rented out the resources to customers who needed them.
In recent years, the demand for training computing resources has surged due to the prosperity of AI development, but due to supply shortages, small startups and universities are struggling to secure computing resources. AI development companies usually use cloud computing resources such as Microsoft and Amazon, but it seems that there are many cases where major cloud companies have reached capacity limits or are not interested in small contracts.
Satoshi Spain's new business focuses on this niche demand, and today it provides computing resources for AI development to startups, universities, and individual developers based in Europe. Alejandro Ibáñez de Pedro, founder of Satoshi Spain, said, 'It is still possible to make money with mining rigs. This is Mining 2.0.'
In addition, Hive Blockchain , a major mining company, reported at its financial results briefing in June 2023 that it is diverting some of its 38,000 GPUs for training large-scale language models. .
In the fourth quarter of 2022, Hive Blockchain rented 500 GPUs as AI computing resources as a beta project and earned $ 230,000 (about 33 million yen). Even if converted to annual revenue, it is about 1 million dollars (about 144 million yen), which is not so conspicuous compared to the scale of Hive Blockchain, but there is a possibility that the scale will further increase in the future.
While cryptocurrency mining uses individual GPUs to compute complex mathematical problems, AI training requires many chips and memories working together to process large amounts of data. Therefore, it is not easy to convert a miner's GPU for AI training. Therefore, some startups buy GPUs that miners have left behind, build their own AI training hardware, and sell resources to AI startups.
According to Vipul Ved Prakash, CEO of cloud computing startup Together, about 20% of the GPUs that miners try to let go can be diverted to AI training. ``Miners were panicking about what to do with their hardware,'' said Prakash, who builds server farms for AI developers with GPUs bought from miners.
The move by miners with spare GPUs to enter the AI computing business is a welcome development for smaller AI startups and research institutes. Demi Guo, founder of AI startup Mellis AI, said that he was using Together's cloud computing because he could not sign a contract with a major cloud computing provider for sufficient capacity. Guo said Together's services are cheaper than those of very mature big companies.
Related Posts:
in Note, Posted by log1h_ik