Image Credit: Bitdeer.com
Key Takeaways
- Bitdeer’s Tydal agreement moves part of its AI strategy from projected capacity and negotiations into a signed, long-duration infrastructure contract.
- The $4.7 billion lease highlights a broader shift from planned AI diversification toward contracted data-center and compute infrastructure.
- Long-term AI and HPC agreements can give miners more revenue visibility than Bitcoin production, which remains sensitive to prices, difficulty and energy costs.
- Existing power access, land and data-center operating expertise are increasingly valuable assets as AI companies compete for grid-connected capacity.
- Riot, TeraWulf, Hut 8 and IREN show that the model is spreading, although companies differ between colocation, data-center leasing and full-stack GPU cloud strategies.
- High capital requirements, customer concentration, technology obsolescence, execution challenges and tighter scrutiny of data-center power consumption remain significant constraints.
Bitdeer Technologies’ latest quarterly disclosure gives the Bitcoin mining industry a more concrete measure of what an AI pivot can become. Its Aug. 10 results highlighted a 16-year lease with Volta for 121 megawatts of critical IT capacity at Bitdeer’s Tydal campus in Norway, representing about $4.7 billion in contractually scheduled payments over the base term assuming full performance.
The agreement, first announced Aug. 4, moves Bitdeer beyond plans to repurpose power assets for artificial intelligence. It also reinforces a broader shift in which miners are trying to become long-duration infrastructure providers for AI and high-performance computing rather than simply adding AI exposure to a crypto business.
The Filing Makes the AI Strategy More Concrete
Bitdeer’s earlier strategy was largely expressed through planned conversions, GPU deployments and negotiations. In September 2025, the company said it expected to operate more than 200 MW of AI computing load by the end of 2026 and described potential annualized revenue above $2 billion under an optimistic full-deployment scenario. The Tydal agreement turns part of that scenario-based case into a signed commercial commitment.
Under the lease, Volta will take 121 IT MW supported by about 133 MW of gross capacity, with the site configured for NVIDIA GPUs serving an unnamed leading AI lab. Bitdeer expects phased commencement between the end of 2026 and the first quarter of 2027. The contract includes 3% annual escalators, reimbursement of electricity costs and an anticipated $1.3 billion letter-of-credit backstop. Bitdeer estimates roughly $500 million of remaining capital expenditure.
That is materially different from saying a mining site could eventually host GPUs. It connects power, construction, and capacity to a customer agreement, giving investors clearer benchmarks for revenue and delivery.
Contracted AI Revenue Could Reshape Miner Economics
Bitcoin mining remains exposed to variables operators cannot control, including Bitcoin prices, network difficulty, block rewards and energy costs. Bitdeer’s second-quarter figures show mining is still the core business: self-mining generated $168.4 million of revenue, compared with $14 million from AI Cloud, while higher electricity and depreciation costs contributed to a companywide gross loss. Bitdeer has previously cited mining difficulty, Bitcoin prices and the halving as factors affecting mining economics.
Long-term AI infrastructure contracts offer a different earnings profile. Lease or colocation revenue can be scheduled over years, often with escalators and power costs passed through to customers. That does not eliminate operating risk, but it can reduce dependence on daily mining economics and improve visibility into future cash generation. Bitdeer’s Tydal structure, for example, provides for electricity reimbursement rather than leaving the company fully exposed to that operating cost.
Miners also bring assets that are difficult to reproduce quickly. Large power interconnections, land, substations, cooling systems and experience running energy-intensive facilities can shorten the route to data-center development. Bitdeer reported almost 3 GW of global electrical and pipeline capacity as of July 31, including sites being evaluated or redesigned for AI use.
An Industry Template Is Emerging
Bitdeer is not creating this trend alone. Riot Platforms has signed a 20-year, 191 MW Rockdale lease that it says represents about $9.1 billion in projected contract revenue, while TeraWulf has a 20-year lease with Anthropic for roughly 401 MW of critical IT load. Hut 8 reported 949 MW of contracted AI data-center capacity and about $26.6 billion of aggregate base-term contract value.
Those agreements suggest power rights and development capability are becoming strategic assets in their own right. IREN has taken a more vertically integrated route through GPU cloud services, showing miners can choose between leasing infrastructure and owning more of the compute stack. Its filings describe large-scale AI Cloud expansion and a $9.7 billion Microsoft contract.
The convergence also intensifies competition for electricity. AI developers, hyperscalers and former crypto miners increasingly seek the same scarce grid connections. Texas Governor Greg Abbott’s August pause on approvals for new data-center grid connections pending an audit shows how energy allocation and local regulation can constrain expansion even when demand is strong.
The Risks Remain Substantial
Contractually scheduled payments are not the same as guaranteed realized revenue. Bitdeer’s Tydal tenant has a no-fee termination right after 10 years, while the proposed credit backstop remains subject to conditions. The project also requires heavy upfront spending at a time when Bitdeer reported $1.8 billion in borrowings and $266 million of second-quarter capital expenditure.
Execution risk is equally important. AI-ready facilities require stronger redundancy, networking and cooling than many mining sites, while miners operating GPU clouds face procurement costs and rapid hardware obsolescence. Bitdeer itself identifies access to GPUs and other critical AI infrastructure components among the risks to its AI business. Customer concentration can also become material when hundreds of megawatts depend on one tenant or end user.
For investors, the next test is whether Bitdeer can turn contracted megawatts into commissioned, occupied capacity on schedule and finance the build without excessive balance-sheet pressure. Additional customer agreements, utilization rates, financing terms and power expansion will show whether the model is repeatable. Bitdeer’s filing does not remove crypto-cycle risk, but it strengthens the case that miners’ power portfolios are being reorganized around a second, increasingly contractual source of demand.
