According to a report from DIGITIMES, over the past five years, Bitcoin miner Core Scientific (CORZ.US) has been quietly shifting its business from mining to the field of artificial intelligence (AI). This field requires a lot of power and workloads to process AI models.

This move is no longer a secret.

On Monday, Core Scientific announced a 12-year agreement with cloud provider CoreWeave to provide infrastructure for use cases such as machine learning. The agreement expands the two companies' existing partnership and will add more than $3.5 billion in revenue over the contract period.

CoreWeave, supported by Nvidia, rents out graphics processing units (GPUs) for training and running AI models, with a valuation of $19 billion in last month's financing round. Core Scientific will provide around 200 megawatts of infrastructure for CoreWeave's operations.

Core Scientific has been mining various digital assets since 2017 and started diversifying into other services in 2019. The company emerged from bankruptcy in January this year.

Core Scientific CEO Adam Sullivan said:"The best way to understand a bitcoin mining facility is that we are essentially the electrical shell of the data center industry."

Sullivan was CEO of Core Scientific when it was in bankruptcy. Since then, the former investment banker has worked with creditors to resolve debt issues and further strengthen the company's non-bitcoin business as it re-enters the public market.

Since its relisting earlier this year, Core has risen more than 40%, but its market cap is around $865 million, far below the $4.3 billion valuation in July 2021.

Bitcoin miners flock to AI field

The debut of ChatGPT by OpenAI in November 2022 triggered an investment frenzy in artificial intelligence models and startups, boosting demand for AI computing and infrastructure. At the same time, other miners such as Bit Digital (BTBT.US), Hive Blockchain (HIVE.US), Hut 8 (HUT.US) and TeraWulf (WULF.US) have been seeking to increase revenue sources, as the so-called Bitcoin halving in April this year cut Bitcoin miners' rewards by 50%.

Many companies have been transforming their large facilities to meet market demands.

"Bitcoin miners generally reside in energy-secure and energy-intensive data centers, and they find these facilities also ideal for AI operations," said James Butterfill, research head at digital asset firm CoinShares.

According to a report from CoinShares, while the capital expenditure required for AI operations is 20 times that of Bitcoin mining, the profit is higher.

CoinShares data shows that 27% of Bit Digital's revenue comes from AI. Hut 8's 6% revenue comes from AI, while Hive, which has data centers in Canada and Sweden, has 4% of its revenue from these services.

Sullivan of Core Scientific said:"While we intend to remain one of the largest and most productive Bitcoin miners, we want to have a diversified business model and more predictable cash flow."

Bitcoin's volatility makes mining a challenging business.

Although Bitcoin has risen more than 150% in the past year to around $69,000, the bear market in 2022 has forced many miners to go bankrupt or shut down.

Turning to AI is not easy.

Turning to artificial intelligence is not as simple as reusing existing infrastructure and machines, as the demand for high-performance computing (HPC) data centers and data networks differs.

In a report on May 30th, Needham analyst wrote: "Almost all the infrastructure that miners currently have, except for transformers, substations, and some switchgear, needs to be levelled and rebuilt to be suitable for HPC."

The equipment used for mining Bitcoin is called an Application Specific Integrated Circuit (ASIC). They are built specifically for cryptocurrency mining and cannot be used for anything else.

Excluding GPUs, Needham estimates that the capital expenditure of HPC data centers is between 8 and 10 million dollars per megawatt, while the capital expenditure of Bitcoin mining sites is usually between 300,000 and 800,000 dollars per megawatt (excluding ASICs).

Is nuclear energy the ultimate goal of AI?

According to Needham analyst, in the next one to two years, the electricity capacity of large listed Bitcoin miners, including mining and HPC business expansion plans, is expected to more than double.

Clean energy is a popular choice as it is often the cheapest source of electricity on many markets. Miners compete in a low-margin industry and their only variable cost is typically energy, so they have an incentive to move to the cheapest source of electricity in the world. An industry report estimates that 54.5% of Bitcoin network power comes from sustainable energy.

The Electric Power Research Institute estimates that by 2030, data centers will account for 9% of total electricity consumption, up from around 4% in 2023. Many people believe that nuclear energy is the answer to meet this demand.

OpenAI CEO Sam Altman said last year that he firmly believes nuclear energy can meet the workload requirements of artificial intelligence.

Altman said: "If we didn't have nuclear energy, I don't think we would be able to meet this goal." "I mean, maybe we could do it with solar and storage. But from my point of view, I think nuclear energy is the most likely and best way."