Mining Giants Pivot to AI: Q2 Earnings Reveal the Stark Reality of Data Center Transitions

Woofun AI reports that the second-quarter earnings landscape for crypto mining entities is defined by a stark dichotomy: the persistent operational noise of Bitcoin mining machines versus the emerging, albeit quieter, financial signals from AI data centers. While MARA and Core Scientific anchor this narrative with divergent strategies, the broader industry is caught in a transitional phase where legacy mining revenues are eroding under price volatility, while new high-performance computing (HPC) infrastructure struggles to translate signed contracts into recognized income. The core tension lies not in the absence of billion-dollar deals, but in the lag between capacity delivery and revenue recognition, exposing the fragility of companies still dependent on digital asset price fluctuations.

MARA's financial performance in the second quarter illustrates the limitations of production growth in the face of asset depreciation. The company mined 2,422 Bitcoins, a modest increase from the 2,358 Bitcoins produced in the same period last year, yet this volume gain failed to offset the broader market downturn. Consequently, MARA's revenue contracted by 27% to $174.9 million. The most significant drag on profitability was the $343 million unrealized loss stemming from the fair value adjustment of its Bitcoin holdings, which contributed to a total net loss of $611.3 million. This data underscores a critical reality: increased hash rate and production efficiency can only partially mitigate the impact of falling Bitcoin prices, leaving miners vulnerable to macroeconomic shifts in digital asset valuation.

Riot Platforms presents an even more pronounced case of margin erosion due to rising operational costs and declining asset values. In the second quarter, Riot mined 1,587 Bitcoins, representing an 11% year-on-year increase in production.

However, its mining revenue fell from $140.9 million to $113.7 million, driven by a drop in the average value per Bitcoin produced from $98,800 to $71,667. Simultaneously, the cost per Bitcoin, excluding mining machine depreciation, rose from $48,992 to $49,912. As a result, the proportion of costs relative to production value surged from 49.6% to 69.6%. This structural shift indicates that without concurrent improvements in scale, machine efficiency, and electricity pricing, the traditional mining model faces severe profitability constraints.

American Bitcoin and Bitdeer offer contrasting perspectives on the relationship between efficiency and expansion. American Bitcoin mined approximately 932 Bitcoins in the second quarter, a 14% quarter-on-quarter increase, generating $67 million in mining revenue, up 8% from the previous period. With a cost per Bitcoin of around $36,500, the company maintained a robust gross margin of nearly 50%, demonstrating that disciplined cost control can sustain profitability even in volatile markets. In contrast, Bitdeer's rapid expansion led to financial strain. The company mined 2,694 Bitcoins, a significant jump from 565 in the same period last year, driving total revenue up 47% to $228.8 million, with $168.4 million derived from self-mined Bitcoin.

However, total costs reached $237.3 million, resulting in a gross loss of $8.5 million and a net loss of $92.3 million. This highlights how aggressive capacity expansion, when outpaced by revenue generation, can quickly erode profitability through electricity costs, depreciation, and expansion expenses.

Core Scientific stands out as the most advanced example of a complete strategic pivot from mining to high-density hosting. In the second quarter, the company reported total revenue of $164.2 million, of which $136.7 million originated from high-density hosting, accounting for 83% of total revenue. Self-mining revenue contributed only $21.5 million, a stark contrast to the previous year when hosting revenue was merely $10.6 million. This dramatic shift confirms that Core Scientific has successfully transitioned its primary income source from selling mined Bitcoin to leasing data center capacity. The company's financial structure now reflects a stable, recurring revenue model based on infrastructure utilization rather than speculative asset appreciation, marking a definitive departure from traditional mining economics.

Woofun AI data shows that TeraWulf and Riot Platforms are in the early stages of recognizing HPC revenue, though their trajectories differ significantly. TeraWulf's second-quarter revenue totaled $44.73 million, with $31.93 million derived from HPC leasing, representing 71% of total revenue, while digital asset revenue accounted for $12.83 million. This follows a trend from 2025, when HPC constituted 90% of TeraWulf's $168.5 million total revenue, including $150 million in HPC-related income and $16.

9 million in initial HPC leasing revenue. Riot's transformation is less mature; its total revenue of $174.2 million, up 14% year-on-year, included $23.2 million from data center operations, comprising $4.9 million in leasing income and $18.3 million from building customer data centers. Mining revenue remained the dominant source at $113.7 million, indicating that Riot's new infrastructure initiatives have not yet reached a scale sufficient to replace legacy mining income.

Cipher Digital and Hut 8 highlight the complexities of construction timelines and mixed revenue streams. Cipher Digital's second-quarter revenue was approximately $24.84 million, entirely from Bitcoin mining, with adjusted EBITDA at negative $30 million and a net loss of $267 million. The company began delivering the first batch of capacity from its Black Pearl project and started charging rent in August, meaning this new revenue stream was not reflected in the second-quarter figures. Hut 8's revenue increased from $41.3 million to $74.9 million year-on-year, with $72.5 million classified under computing services.

However, this category includes ASIC computing, AI cloud services, and traditional cloud services, preventing a direct attribution to AI revenue alone. Hut 8's net loss of $177.1 million, largely driven by $138.6 million in unrealized digital asset losses, further obscures the true profitability of its diversified service offerings.

A prevalent misconception in the industry is the conflation of long-term contract values with recognized revenue. Core Scientific reported renting out approximately 1.1 GW of capacity, corresponding to potential contract revenue exceeding $24 billion, yet its recognized hosting revenue in the second quarter remained at $136.7 million. Similarly, TeraWulf signed a 20-year lease agreement with Anthropic after the quarter, with an initial contract value of around $19 billion, but its HPC leasing revenue for the period was only $31.9 million.

Riot also signed a 191 MW data center lease agreement with an initial value of approximately $9.1 billion, while its data center revenue was $23.2 million. These figures are not contradictory; they reflect the accounting reality that revenue is recognized quarterly upon capacity delivery and rent commencement, not upon contract signing. Project delays, construction cost changes, and financing arrangements all influence the timing of revenue recognition, making it essential to distinguish between signed contracts, delivered capacity, and recognized income.

Net profit metrics are further distorted by accounting nuances, particularly fair value adjustments. Core Scientific's net loss of $1.1553 billion was primarily driven by changes in the fair value of options, rather than operational cash flow. Cipher's net loss of $267.5 million included a $150.5 million loss due to similar fair value adjustments. MARA's losses were heavily influenced by revaluations of Bitcoin prices, while Bitdeer's gross loss indicated that operational costs had exceeded revenue, a fundamentally different type of financial distress. These accounting items obscure the underlying operational performance, requiring investors to look beyond net profit to assess the true health of each company's transition strategy. The divergence between operational cash flow and reported net income highlights the need for a more nuanced evaluation of mining companies' financial statements.

Keel Infrastructure, formerly Bitfarms, represents the most extreme case of exit and transition. The company shut down its Bitcoin mining operations in the United States, specifically in the Moses Lake area, in April 2026. Its second-quarter revenue was approximately $30.43 million, a 50% year-on-year drop, with adjusted EBITDA at negative $23.7 million. Keel has chosen to become an HPC infrastructure developer, but its new business has not yet reached a scale sufficient to replace mining revenue.

This quarter's earnings reports reveal that the industry is no longer monolithic; some companies like American Bitcoin still focus on hash rate expansion, while others like Core Scientific and TeraWulf have established significant hosting revenues. Companies like Riot and Cipher are in the middle of delivering new projects, and Keel is navigating a complete business model overhaul. The future financial performance of these entities will depend on their ability to secure stable electricity, deliver data centers on time, and convert long-term contracts into actual, recognized revenue.