A major player in Bitcoin mining, Core Scientific, has entered into a strategic partnership with AMD, one of the world's leading developers of processors and graphics chips. Under the terms of the agreement, Core Scientific will provide AMD with up to 2.5 gigawatts (GW) of its data center capacity. These resources will be used to deploy and support client artificial intelligence (AI) systems.
Shift in Focus: From Mining to AI
This deal is a clear indicator of how mining companies are adapting to new market realities. Instead of relying solely on the volatile cryptocurrency market, Core Scientific is diversifying its business, transforming its energy-intensive data centers into hubs for high-performance computing (HPC) and AI workloads. AMD, in turn, gains access to infrastructure capable of providing stable power supply and cooling for the most demanding AI chips, including the Instinct accelerator line.
Agreement Details and Potential Impact
The allocation of up to 2.5 GW of capacity is not a one-time delivery but a long-term contract that, in my estimation, will be rolled out in phases over several years. Core Scientific is known for its large-scale facilities, previously optimized for Bitcoin mining. These same facilities will now be upgraded for AI computing, requiring significant investment in liquid cooling systems and high-speed networks.
For the market, this signals that competition for energy resources between traditional mining and the AI sector will only intensify. Core Scientific is betting that the profitability from leasing capacity for AI could be higher and more stable than from cryptocurrency mining, especially during periods of declining hash rate or network difficulty.
Expert Opinion
This move by Core Scientific is pragmatic and forward-looking. Integrating AI infrastructure into the business of mining companies is not just a trend but a new reality. Those who can effectively repurpose their assets and secure contracts with tech giants will gain a significant competitive advantage. However, it is worth noting that transitioning to servicing AI workloads requires a completely different level of engineering expertise and capital expenditure, which could become a barrier for less prepared players.