Hive Blockchain Promotes the Privacy of AI Models Using its GPUs

• Hive still uses 38,000 GPUs from Ethereum mining.
• Probabilistic calculations help large language models understand and create human language.

In order to support its shift to artificial intelligence (AI), Hive aims to provide corporate training on its graphics unit processing (GPU) fleet.

Hive Blockchain Attracts Customers to its AI Model

With more privacy than competitors like OpenAI’s ChatGPT, bitcoin mining company Hive Blockchain (HIVE) wants to enable clients to train big language AI models in its data centers, stated in a Friday conference call with analysts.

Companies are increasingly aware they shouldn’t provide private customer information to an organization like OpenAI that uses a public large language model (LLM). According to Aydin Kilic, CEO and President of the mining company, “What we want to offer at Hive through Hive Cloud is privacy where companies can have a service agreement in place, ownership of their data, and privacy while still running artificial intelligence (AI) compute workloads on our bank of graphics processing units (GPUs).

Miners Turning Towards AI

As a result of the interference that mining economics have caused with their profitability, miners are increasingly turning their attention to artificial intelligence. This comes at a time when the AI industry is seeing a surge in interest from investors. 

However, it is yet unknown whether miners will be able to compete with large technology companies like Google and Amazon Web Services, which have the benefits of economies of scale as well as decades of experience operating high-quality customer-facing data centers.

Large language models use probabilistic calculations to comprehend and produce human language. They are often trained on graphics processing units (GPUs), a sort of semiconductor-based electrical circuit that was first employed for picture processing but has since been shown to be effective at handling AI workloads.

Working of HIVE GPUs

38,000 GPUs are still in use by Hive from the time when it was mining Ethereum. Some of them have been used to mine alternative cryptocurrencies, while others are placed in its cloud service or may be rented as a service.

During the results call, the company said that their GPU run rate is anticipated to be $1 Million yearly. In a press statement announcing the company’s annual and quarterly profits, the firm’s chairman, Frank Holmes, said that “500 GPU cards generated $230,000 in revenue this quarter.” On March 31, 2023, Hive’s fiscal year came to a conclusion.

Hive reported $106.3 million in revenues for the fiscal year that ended on March 31, 2023, with a gross operating margin of $50.4 million, or 47% of revenue. At a gross operating margin of 78% of sales, that is roughly half of the historical revenue the company reported in FY2022 of $211.2 million.

Overall, the company recorded a net loss of $236.4 million for FY2023, which it highlighted includes a number of non-cash charges, including $81.7 million in depreciation, $70.4 million in equipment impairment, and $27.3 million in impairment of deposits. In comparison, the company recorded a fiscal year 2022 net income of $79.6 million.

APLD Made a Similar Move as HIVE

Applied Digital (APLD) has signed a deal to host cloud computing workloads for artificial intelligence (AI) in its data centers. The contract may be valued up to $460 Million over a 36-month period.

Following the news of the second AI hosting partnership in as many months, the Texas-based cryptocurrency miner’s stock price increased by as much as 17% on Nasdaq.

Like other cryptocurrency miners, Applied Digital has been exploring other sources of income to supplement income in the face of sluggish cryptocurrency markets. AI, cloud computing, and other high-performance computing applications are all potential future developments.

The May arrangement, according to Applied Digital, was worth up to $180 Million over the next two years. For its AI and cloud services, it has joined forces with hardware manufacturer, Supermicro.