Bitcoin mining companies are repositioning themselves as potential AI infrastructure providers, using land, substations and grid connections built for energy-intensive mining. But the value of those assets depends less on headline contract totals than on the cost of conversion, the timing of revenue and the claims held by customers and lenders.
According to CryptoSlate, Bitcoin miners have announced more than $100 billion in AI-related contracts, yet only about 550 megawatts (MW) of more than 4 gigawatts under contract was generating billing revenue in September. The gap highlights the difference between securing power and operating a finished AI data center.
Why mining sites appeal to AI developers
Bitcoin mining facilities already occupy industrial sites connected to large amounts of electricity. Their specialized ASIC mining machines cannot train AI models, but the surrounding infrastructure can provide a starting point for data-center developers facing lengthy grid-connection processes.
That starting point is not the same as a completed AI facility. AI data centers require denser cooling systems, high-speed networking, redundant power arrangements and stronger uptime requirements than conventional mining sites. They may also need major construction work and new financing before customers can begin paying.
The distinction matters when companies report power figures. Electricity already flowing to a site, capacity approved for future use and a position in a utility connection queue are different things. Treating them as one number can make an infrastructure portfolio appear closer to producing AI revenue than it really is.
SWI Group said in June that Genesis Digital Assets, which it took control of earlier this year, had more than 1.3 GW of energized and approved connections across 15 facilities. Its half-year results later referred to approximately 1.2 GW of secured grid connections, without reconciling the figures or providing a site-by-site split between usable capacity and power requiring additional work.
SWI announced an €800 million capital increase on October 12 to expand its AI infrastructure business. On October 26, its SWI Compute unit disclosed about $1 billion of NVIDIA equipment orders and its first customers, with some contracts still being finalized. It also announced a 13 MW infrastructure partnership in Spain.
The strategy is broader than simply leasing powered buildings. SWI intends to sell computing capacity itself, which could provide access to GPU-related revenue but also requires spending on equipment and creates exposure to utilization and hardware obsolescence. The company has not disclosed which Genesis sites will host the equipment, how much of the reported grid capacity can be converted economically, or when those sites may begin generating AI revenue.
Conversion costs can change the economics
The transition is expensive. CoinShares estimated in its September mining report that converting mining infrastructure to AI-grade facilities costs roughly $8 million to $15 million per MW, compared with $700,000 to $1 million per MW for conventional mining infrastructure.
Power consumption also needs careful interpretation. At a power usage effectiveness ratio of 1.25, a facility drawing 100 MW from the grid would deliver about 80 MW to computing equipment, with the remainder supporting cooling and power distribution. A 100 MW grid connection therefore cannot automatically be compared with 100 MW of usable computing capacity.
Core Scientific’s accounts show why the old mining business has become less attractive. In the first quarter, it wrote down $266.5 million of mining equipment and infrastructure. Mining machines accounted for $151.6 million of the charge and mining infrastructure for $114.9 million. The company’s mining operation was losing money at the gross-profit level, while its newer AI hosting activity produced a positive gross margin.
Core Scientific’s second-quarter colocation revenue reached $136.7 million, with 437 MW billing customers by mid-July. Of that revenue, $98.8 million came from license fees and $35.1 million from power charges passed through to customers. Because the electricity reimbursement appeared as an equivalent expense, it increased reported revenue without creating gross profit.
CoreWeave contracts reveal who carries the cost
Agreements between Core Scientific and AI cloud provider CoreWeave show how the economics are divided. Most covered sites begin with a base license fee of $100 per kilowatt per month, rising 3.5% annually, while CoreWeave is responsible for utility costs.
CoreWeave also funds or reimburses certain approved construction expenses. For specified landlord-owned equipment, it can receive credits against future base fees worth as much as $1.5 million per MW. Those credits can reduce the relevant rent payment by half until used.
At the initial $100 rate, one MW would generate a contractual base fee of $100,000 a month. A 50% construction credit would reduce the payment to $50,000. If the full $1.5 million allowance applied, it would take 30 months of such credits to recover the amount, before rent increases and other contract provisions.
The arrangement helps Core Scientific fund construction without raising every dollar itself, but it also means that nominal contract value does not equal early cash receipts. The company’s filing cites prepayments, construction credits and straight-line revenue recognition as reasons accounting revenue can differ from cash collected.
Core Scientific recorded $954.2 million in capital expenditure during the first half of 2026, including $180.9 million funded by CoreWeave. A financing subsidiary also issued $3.3 billion of secured notes at 7.75% in May. At the original principal, the annual coupon cost is approximately $255.8 million before repayments.
The contracts include daily late-delivery credits of $20,000 and then $30,000 in specified circumstances, while some delays can give CoreWeave termination rights. Core Scientific also provided an uncapped completion guarantee for specified financed projects if available construction funds prove inadequate. The company keeps ownership and potential long-term income, but assumes substantial construction and financing obligations.
Market picture
Moneyiar market data at 12:48-12:49 on November 1, 2026, showed Bitcoin at $77,768.26, up 0.46% over 24 hours but down 3.19% over seven days. Celestia was $0.4797, up 0.06% daily and down 8.72% weekly. Compound stood at $20.59, up 1.32% in 24 hours and down 9.79% over seven days, while OriginTrail was $0.3539, up 3.58% daily and down 4.09% weekly.
The broader lesson from the mining-to-AI shift is that a power connection is an important asset, but not a completed business. Construction providers, equipment suppliers, tenants and creditors may all have contractual claims before shareholders receive the benefits of a converted site. The outcome will depend on how effectively companies turn secured electricity into operating capacity while managing those competing obligations.
