Crypto companies that have overloaded GPUs in data centers for coin mining are now considering whether to sell or repurpose unused hardware to the exploding artificial intelligence market.
Powered by 38,000 Nvidia GPUs, Hive Blockchain repurposes hardware for high-performance computing and artificial intelligence applications. The company began rethinking what to do with its GPUs (still used to mine Bitcoin and Ethereum) after the cryptocurrency market crash.
Cryptocurrency companies are starting to realize that GPUs can generate more revenue for AI and cloud operations. Hive successfully ran a pilot program to dedicate its GPUs to AI and high-performance computing.
“We have a lot of work to do to fully utilize our GPU cards, but we are very pleased that a beta project using just about 500 GPU cards generated $230,000 in revenue this quarter. said Frank Holmes, Executive Chairman of the company. last month.
Cryptocurrency mining firms are operating in a volatile market as the Bitcoin price crashes in late 2021. Hive noted that bitcoin prices have fallen an average of 49% since last year.
The coin price crash coincided with software changes that made mining less reliant on GPUs.
Mining companies especially felt a lot of pressure after Ethereum’s software switched from the old proof-of-work system to a proof-of-stake model. A new system implemented to reduce the energy consumed in data centers did not require the fastest chips to earn Ethereum. This has reduced our reliance on GPUs for mining coins.
But GPU versatility is still an asset to Hive. Demand for GPUs has skyrocketed with the advent of artificial intelligence, and Nvidia’s latest GPUs are in short supply. Hive is powered by his old Nvidia GPU, which is also in demand for AI inference.
The volatility of the cryptocurrency market has forced government services company Sysorex to reconsider what to do with its 12,000 GPUs as well. Its crypto-focused subsidiary TTM pulled out of the mining business after Ethereum switched software models.
As a result, Sysorex doesn’t know what to do with its precious GPU. The cryptocurrency business model is broken and repurposing GPUs for other applications is not feasible as it requires additional upfront investment in data centers, Sysorex said in a U.S. stock exchange on June 30. said in a submission to the Commission.
“TTM Digital is currently exploring alternative uses and sales opportunities for its graphics processing unit (“GPU”) assets and data center located in Lockport, NY,” wrote Sysorex.
Miners used to be a headache for Nvidia and AMD. Both companies have experienced unexpected increases and decreases in revenue from GPU sales to cryptocurrency miners, which has been difficult to predict. Miners also devoured the limited supply of GPUs, paying premiums during chip shortages.
Chip makers eventually made changes to make GPUs more aimed at the core gamer demographic and less attractive to miners. Additionally, Intel retired a Bitcoin ASIC called BlockScale earlier this year.
A shortage of GPUs and an over-reliance on Nvidia GPUs for AI computing is also forcing customers to source alternative chips from other chip makers. Intel and AMD have introduced GPUs for AI, but are not shipping chips in large numbers at this time.
AI tends towards large language models like OpenAI’s ChatGPT and Microsoft’s BingGPT, but these require large GPU-powered clusters. But open-source developers and companies like MosaicML have also introduced open-source models that can also be run locally on laptops and desktops (i.e. run trained models on GPU clusters on new data) for inference computing. doing. Intel also included AI accelerators in the new CPUs that provide inference processing.
