Core Scientific's AI Pivot Yields $80M Profit Despite 56% Mining Margin Collapse

Woofun AI reports that Core Scientific’s strategic pivot from Bitcoin mining to AI hosting accelerated in the second quarter, resulting in a stark divergence between its legacy and emerging business lines. While the company continues to operate as a longtime Bitcoin miner, the financial data reveals that its conversion of sites for AI computing is generating significantly higher profitability, even as the traditional mining segment struggles with severe margin compression.

The mining segment’s financial performance was characterized by a negative 56% self-mining gross margin, a figure that underscores the deteriorating economics of its core legacy business. This margin calculation is distinct from a disclosed spot-Bitcoin breakeven or a cash-production-cost estimate, as the cost of revenue included substantial fixed and variable expenses. Specifically, the segment incurred $17.9 million in power fees and $9.9 million in depreciation and other operating expenses. Consequently, the reported margin cannot be reduced to the price at which the machines cover electricity alone, indicating that the operational burden extends well beyond basic energy consumption. This structural cost disadvantage highlights why the mining operation is no longer the primary driver of value creation for the firm.

In sharp contrast, the high-density colocation segment, which provides powered data-center capacity for AI customers, delivered robust financial results. The segment produced $136.7 million in revenue and $80.0 million in gross profit, achieving a 59% margin.

Notably, this $80.0 million gross profit exceeded Core Scientific’s $70.0 million consolidated total for the quarter, as losses from the mining and other segments pulled the companywide figure lower. This dynamic illustrates the immediate economic benefit of the new business model. Core Scientific states it is repurposing its remaining mining facilities for high-density colocation "as circumstances allow," a strategy that the Q2 loss strengthens economically, although the company did not identify the quarter as a trigger for compulsory conversion.

Per Woofun AI, the operational wind-down of the mining business is being managed to offset contractual power costs during the transition. CFO Jim Nygaard explained that the company was operating mining primarily for this offset purpose. He noted that Core Scientific ended June with nearly 30% fewer miners online than at the end of the first quarter and was self-mining at only two sites.

Meanwhile, the colocation capacity expanded, with the company reporting 395 megawatts of billing colocation capacity at quarter-end and 437 MW by mid-July. The latter figure represented approximately $635 million in average annualized colocation GAAP revenue, signaling rapid scaling in the new segment.

Despite this growth, the current operational footprint remains well below the roughly 1.1 gigawatts of leased customer power capacity tied to more than $24 billion of potential contracted revenue. The AMD relationship is anchored by 15-year agreements covering about 530 MW across five sites and more than $14 billion of potential base contracted revenue. A broader relationship could support up to 2.5 GW, but that figure is prospective, not built or billing capacity. This gap highlights how much of Core Scientific’s AI story still depends on future conversion and delivery.

Furthermore, the company’s $1.16 billion net loss overstates operating damage, as it was primarily driven by a $1.05 billion fair-value expense for warrants and contingent value rights as the stock price rose, rather than core operational failures.

The quarter does not prove that AI conversion is forced, but it clarifies why mining is losing its claim on the company’s power and sites. One segment produced a negative gross margin while the other generated more gross profit than Core Scientific recorded in total. This marks a decisive shift in the company’s value proposition, where the viability of the business model now hinges on the successful execution of its AI infrastructure pipeline rather than the fluctuating profitability of Bitcoin mining.