This does not mean the mining model is no longer viable; rather, scale, machine efficiency, and electricity costs must work together. In the second quarter, American Bitcoin mined approximately 932 bitcoins, a roughly 14% increase quarter-over-quarter; mining revenue reached about $67 million, up approximately 8% quarter-over-quarter. Its cost per mined bitcoin is around $36,500, with a gross margin nearing 50%.
Bitdeer, however, revealed another side: while hashing power and output can expand rapidly, profits do not necessarily follow suit. In the second quarter, the company mined 2,694 bitcoins, compared to 565 in the same period last year; total revenue increased 47% year-over-year to $228.8 million, with $168.4 million coming from self-mining. However, the company’s cost of revenue for the quarter reached $237.3 million, resulting in a gross loss of $8.5 million and a net loss of $92.3 million. Focusing solely on output and revenue can easily obscure the fact that electricity, depreciation, and expansion costs have already exceeded current revenue.
AI revenue has already emerged, but companies are not starting from the same line.
What is truly transforming the industry is that some mining companies have shifted from "selling the bitcoins they mine" to "renting out power and data center space."
Core Scientific is the company with the most noticeable progress. In the second quarter, the company reported total revenue of $164.2 million, with high-density hosting revenue reaching $136.7 million, accounting for approximately 83% of total revenue; self-mining revenue was only $21.5 million. In the same period last year, Core Scientific’s hosting revenue was just $10.6 million. This indicates that its primary revenue source has already shifted from mining to data center hosting—not merely an announced future project.
TeraWulf has undergone a similar structural shift. In the second quarter, the company generated $44.73 million in revenue, with $31.93 million coming from HPC leasing, accounting for approximately 71%, and $12.83 million from digital assets. In contrast, in 2025, TeraWulf’s total revenue of $168.5 million was still dominated by mining at 90% (approximately $150 million), though HPC leasing revenue was generated for the first time at $16.9 million. The company has clearly stated that capital allocation and operational focus will primarily center on HPC data centers, with some existing mining infrastructure being repurposed.
Riot’s transition is still in its early stages. Its total revenue for the second quarter was $174.2 million, a 14% year-over-year increase; of this, data center revenue amounted to $23.2 million, while mining revenue still reached $113.7 million. Notably, Riot’s data center revenue includes $4.9 million in leasing income and $18.3 million in customer data center construction income—both of which are genuine revenues recognized in the current quarter—but they are still insufficient to replace the mining business.
Cipher Digital reminds the market that "beginning construction" and "having generated HPC revenue" are two different things. The company's second-quarter revenue was approximately $24.84 million, all from Bitcoin mining, with an adjusted EBITDA of -$30 million and a net loss of $267 million. Cipher only began delivering the first capacity of its Black Pearl project in early August, so this revenue has not yet been reflected in the second-quarter results.
Hut 8's second-quarter revenue increased from $41.3 million in the same period last year to $74.9 million, with $72.5 million attributed to computing services. However, this category includes ASIC computing, AI cloud, and traditional cloud services, so the $72.5 million cannot be directly labeled as AI revenue. Nonetheless, Hut 8 still reported a net loss of $177.1 million for the second quarter, of which $138.6 million stemmed from unrealized losses on digital assets.
AI large orders flood the screen; revenue realization still requires time
The most common misunderstanding during the transition of mining companies is confusing the total value of long-term contracts with realized revenue.
Core Scientific disclosed that its leased customer power capacity is approximately 1.1 GW, corresponding to over $24 billion in potential contract revenue, yet its recognized托管 revenue for the second quarter remained at $136.7 million; TeraWulf signed a 20-year lease agreement with Anthropic after the quarter, with an initial contract value of approximately $19 billion, but the company’s HPC leasing revenue for the second quarter was only $31.9 million; Riot signed a 191 MW data center lease agreement after the quarter, with an initial value of approximately $9.1 billion, while its data center revenue for the second quarter was $23.2 million.
These figures are not contradictory. The total contract value represents the potential revenue over the entire lease term and is typically recognized quarterly in financial statements after the data center is built, capacity is delivered in phases, and rental charges begin. Delays in the project, changes in construction costs, financing arrangements, and customer performance can all affect the actual recognition timeline. Therefore, when comparing AI businesses of mining companies, it’s essential to distinguish between three key metrics: how many contracts have been signed, how much capacity has been delivered, and how much revenue was recognized in the current quarter.
Net profit cannot be read in isolation from accounting items. For example, Core Scientific reported a net loss of $1.1553 billion in the second quarter, primarily due to changes in the fair value of warrants; Cipher incurred a net loss of $267.5 million, including a $150.5 million fair value loss on warrants; MARA’s loss was influenced by the revaluation of Bitcoin prices. In contrast, Bitdeer reported a gross loss for the quarter, reflecting that the cost of revenue exceeded revenue—a fundamentally different situation.
Mining companies are splitting into three types of firms.
The Q2 earnings report shows that publicly traded mining companies can no longer be measured by the same standards.
American Bitcoin companies continue to focus primarily on expanding hashing power, increasing production, and lowering per-unit costs; Core Scientific and TeraWulf already have a significant portion of their托管 or HPC revenue reflected in current financial statements; Riot, Cipher, and other companies are in the intermediate stage of gradual project delivery.
Keel Infrastructure has gone further. The company, formerly known as Bitfarms, has shut down its U.S. Bitcoin mining operations, reporting second-quarter revenue of approximately $30.43 million—a 50% year-over-year decline due to the drop in Bitcoin price and the closure of its cryptocurrency mining activities in Moses Lake, U.S., in April 2026; adjusted EBITDA stood at negative $23.7 million. The company has chosen to transition into an HPC infrastructure developer, but its new business has not yet generated revenue scales sufficient to replace mining income.
Therefore, what truly matters this quarter is not whether mining companies are talking about AI, but how far they’ve progressed: some are still boosting output through mining hardware, others have begun collecting monthly data center rents, and some are navigating the transition period where old revenues have disappeared but new ones haven’t yet materialized.
The mining rigs are still running, but what will determine the next phase of financial performance is who has access to stable electricity, who can deliver data centers on time, and who can truly turn a long-term contract into current revenue.
