Core Scientific transitions from Bitcoin mining to AI infrastructure

The exit from Bitcoin mining

The transition from Bitcoin mining to AI infrastructure defines the current turnaround for Core Scientific. The company plans to exit all self-mining operations by December 31, 2026. This decision follows a period of intense volatility in the digital asset market. I see a company that is deliberately shrinking its mining operations to prioritize high-density colocation services. The company reported that revenue from its self-mining segment fell to $12.3 million in the third quarter of 2026. This is a 79% drop compared to the same quarter in the previous year. In the same period, colocation revenue reached $215.4 million. This amount is a nearly 60% increase from the second quarter of 2026. The company has already reduced its mining fleet by almost 50% since the end of 2025. This shift allows the company to reallocate power and floor space to long-term customer agreements. Management, including CEO Adam Sullivan, CFO Jim Nygaard, COO Matt Brown, and SVP Jon Charbonneau, intends to end all self-mining by the end of 2026. This timeline leaves the company with zero mining revenue after the new year. The company emerged from Chapter 11 bankruptcy in January 2024 and has since worked to rebuild its balance sheet. Net loss for the second quarter of 2026 was primarily driven by the change in fair value of warrants, which reflects appreciation in the stock price.

AI capacity and contracted revenue

AI capacity contracts provide a massive new revenue foundation. The partnership with AMD supports up to 2.5 GW of leasable capacity through 15-year agreements for 530 MW across five sites, which could generate more than $14 billion in potential base contracted revenue. The deal includes both AMD and a neocloud counterparty. The company also maintains a long-term contract with CoreWeave for approximately 590 MW. In 2024, the company announced that these 12-year agreements were expected to generate more than $3.5 billion of cumulative revenue over their terms. Total leased customer power capacity is approximately 1.1 GW. This capacity equals more than $24 billion of potential contracted revenue. The company reported billing capacity of 485 MW as of November 5, 2026. This is an increase from the 437 MW reported in mid-July 2026. I think the heavy reliance on a few large tenants like AMD and CoreWeave creates a unique risk profile. The jump from 485 MW of billing capacity to the multi-gigawatt potential of the AMD deal requires massive construction and power procurement.

Capacity Metric Value
Total Leased Customer Power Capacity 1.1 GW
AMD Contracted Capacity 530 MW
CoreWeave Contracted Capacity 590 MW
Reported Billing Capacity (Nov 2026) 485 MW
Potential AMD Base Revenue $14 billion+
Total Leased Contracted Revenue Potential $24 billion+

Financial divergence and asset repurposing

The financial performance of the company shows a massive divergence between its legacy mining and its new AI services. In 2025, the company generated $229.2 million from digital asset self-mining. That same year, colocation revenue was $65.4 million, which was a $41 million increase over 2024. By the third quarter of 2026, the situation flipped entirely. Colocation revenue for the third quarter was $215.4 million. The self-mining segment brought in only $12.3 million. The colocation segment reported a 57% adjusted EBITDA margin in the third quarter. This margin is strong because the company repurposes its existing mining facilities for high-density colocation. In 2025, the company had a total mining hash rate of 17.9 EH/s with 151.4 miners, which was down from 20.1 EH/s and 171.1 miners in 2024. The company’s facilities are located in Alabama, Georgia, Kentucky, North Carolina, North Dakota, Oklahoma, and Texas. In March 2026, the company sold approximately $175 million worth of Bitcoin holdings to accelerate its AI infrastructure expansion. The company also took out a billion-dollar loan to build new data centers in several US states. I believe the company’s ability to repurpose these sites gives it a speed advantage. However, the transition has costs. The company lost $41.9 million in 2026 because it terminated a mining chip supply agreement with Block/Proto. This loss is a direct consequence of the strategic pivot. The company also reported a net loss of $288.6 million for the year ended December 31, 2025, which was a significant improvement from the $1.4 billion loss in 2024.

Infrastructure and power access

Access to power and specialized infrastructure dictates the success of high-performance computing. Core Scientific builds large scale, purpose-built data centers for high-density workloads. These facilities provide the cooling and power connections that AI customers need. The company operates 10 data centers and holds enough cash to cover the 1 GW essential IT load for its current and future sites. It is also expanding its Denton, Texas, campus by 72 MW. This expansion aims to increase mining hashrate by 20 exahash. The Denton site has a total contracted power of approximately 300 MW. The company also operates a Pecos, Texas, facility with 250 MW of total contracted power. For example, the Pecos site operates 71 MW of Bitcoin mining across two sites. The company uses utility power from providers like Alabama Power, Duke Energy, and Texas New-Mexico Power. Specifically, the Dalton, Georgia, facility uses 195 MW from Dalton Utilities. Management expects to move all existing capacity to colocation within three years. The company also maintains a long-term contract with CoreWeave that covers 590 MW. You should watch if the company can meet its contractual obligations as it scales. Can the company deliver the full 530 MW of AMD capacity by the 2028 target?

Newsletter