A megawatt leased to an AI tenant now commands a different price on Wall Street than a megawatt sitting in a Bitcoin miner’s pipeline, and the distance between the two has become the central pricing question for the entire sector.
VanEck’s latest framework for valuing publicly traded miners shows that companies with signed AI and high-performance computing leases trade at more than 10 times gross energy output, while miners with little or no contracted capacity trade at roughly 2 to 6 times that metric.
Investors have started treating leased megawatts as a distinct, more valuable asset class than mined Bitcoin or unsold power capacity.
MetricVanEck figureWhy it mattersMiners with signed AI/HPC leasesAbove 10x gross energized powerWall Street is assigning a premium to contracted AI capacityMiners with little or no contracted capacityRoughly 2x–6x gross energized powerPipeline alone is worth much less than signed leasesDelivered AI/HPC capacity~25% of leased capacityMost contracted capacity still has to be built and deliveredNear-term funding shortfall~$50BThe sector needs major capital before leases become cash flowLong-term capital need if pipelines convert~$221BThe AI pivot could become an infrastructure-scale financing cycleA
The premium is arriving before the capacity
VanEck puts delivered AI and HPC capacity across the peer group at only about 25% of what has been leased. Wall Street is paying for contracts today and for construction outcomes the sector has not yet delivered.
The near-term funding shortfall for that construction totals roughly $50 billion across the group, with long-term capital needs climbing toward $221 billion if the full pipeline of announced projects ultimately converts into built sites.
VanEck’s valuation model assumes a baseline net operating income of about $1.5 million per megawatt for AI and colocation sites and applies an enterprise value multiple of 15 times that figure.
The model also offsets the result against greenfield construction costs of roughly $10 million per megawatt, climbing to about $12 million for projects further out as construction inflation compounds.
A single megawatt implies a gross enterprise value near $22.5 million, against a pre-financing value of about $12.5 million after capex, before any probability discount for delivery risk or financing costs is applied.
InputAssumptionImplied valueNet operating income per MW~$1.5MStarting cash-flow baseEnterprise value multiple15xConverts NOI into asset valueGross enterprise value per MW$1.5M × 15~$22.5MGreenfield construction cost~$10M/MWBaseline capex deductionPre-financing value after capex$22.5M – $10M~$12.5MFurther-out project capex~$12M/MWLower implied equity value if costs riseMain sensitivityCapex, timing, tenant qualitySmall changes can materially alter shareholder upside
Pushing the capex per megawatt up by a few million dollars, or stretching the delivery timeline by a year, and the equity value attached to that megawatt moves by a proportionally large amount.
VanEck’s framework treats a megawatt leased to an investment-grade hyperscaler as supportable at an effective cost of capital between 6% and 10%. A similar megawatt leased to a smaller GPU cloud tenant can warrant a discount rate above 10%, the cost of capital growing directly with tenant risk.
A signed lease and an energized megawatt carry different values once the tenant’s balance sheet is factored in. The same power, sold to a weaker counterparty, commands a smaller premium.
Financing the shortfall without giving away the upside
Closing a $50 billion near-term shortfall pulls miners toward financing tools drawn from infrastructure and project finance.
Project finance and debt bring fixed obligations onto balance sheets built around volatile mining margins. Bitcoin treasury sales convert an asset some miners spent years accumulating into construction capital, undercutting the original thesis that drew Bitcoin-focused investors into the stock in the first place.
Strategic partnerships and tenant prepayments offer a softer path, but they typically come with terms that shift a portion of the AI-era upside away from existing shareholders and toward whichever partner supplies the capital.
The International Energy Agency projects that global data center electricity consumption will roughly double from about 485 terawatt-hours in 2025 to around 950 terawatt-hours by 2030, with AI-specific data center consumption tripling over the same period.
McKinsey estimates that global data center spending could reach about $7 trillion by 2030, with roughly $5.2 trillion directed toward AI-capable facilities.
KKR’s recently launched $10 billion AI infrastructure venture with Nvidia, and Vistra shows large financial institutions treating power-backed AI capacity as its own asset class, with capital scaling at a pace that matches the size of the opportunity miners are chasing.
Bitcoin’s shadow hasn’t lifted
The market continues to price miners based on Bitcoin’s daily swings, even as VanEck’s framework describes a business model migrating toward AI leases.
The peer group’s average one-year weekly beta to Bitcoin is near 1.05, meaning the typical mining stock still moves in near lockstep with Bitcoin’s price, even as its underlying cash flow story shifts toward AI leases.
Meaningful Bitcoin treasury exposure, the kind that would justify that beta, is concentrated in a handful of names.
Company / groupBTC holdings as % of market capWhat it suggestsMARA~51%Still meaningfully tied to Bitcoin treasury valueCLSK~24%BTC exposure remains materialRIOT~11%Some BTC balance-sheet linkageHUT~7%Limited but visible BTC exposureMost other peers~1% or lessBTC beta may overstate actual balance-sheet exposurePeer-group average beta to BTC~1.05Stocks still move almost one-for-one with Bitcoin
MARA holds Bitcoin worth about 51% of its market cap, CLSK around 24%, RIOT near 11%, and HUT roughly 7%, while most peers hold Bitcoin at 1% or less of their market cap.
AI-focused winners can trade too cheaply during a Bitcoin selloff, while pipeline-heavy laggards can trade too richly whenever Bitcoin rallies.
VanEck’s governance scorecard evaluates insider ownership, management KPIs, executive compensation structure, leadership tenure, and related-party transactions, and finds no company in the group scoring close to a perfect mark, with HIVE and BTDR ranking lower on the relative scale.
Funding tens of billions of dollars in AI infrastructure requires investors to trust management teams with capital budgets several orders of magnitude larger than anything a mining-era balance sheet previously demanded.
Governance gaps carried little consequence in a hash-rate business, and real weight in one that sells power to hyperscalers under long-dated contracts.
Two paths from contract to cash flow
A bull case for the sector is that miner valuations migrate toward the framework already used for data-center REITs and infrastructure landlords.
Hyperscaler demand for power-dense, interconnection-ready sites stays intense, financing markets open up for creditworthy projects, and the miners furthest along in construction begin reporting delivered megawatts and recurring lease revenue.
Multiple-on-delivered capacity holds near or above the 10x level that VanEck already observes, and the premium the market assigned early is validated by the cash flow that eventually follows.
A bear case has the funding shortfall resolved through dilution, as construction costs climb past the $10 million-per-megawatt baseline due to rising labor, equipment, and grid interconnection expenses.
Debt gets priced for a sector with limited operating history as an infrastructure landlord, pushing miners toward equity issuance or Bitcoin monetization to bridge the shortfall before AI revenue materializes.
Shareholders fund the buildout, and a meaningful share of the eventual upside flows instead to lenders, strategic partners, or the buyers of newly issued equity who priced their entry after the dilution.
The test that decides which case plays out has nothing to do with the size of a miner’s next AI announcement.
It comes down to delivered megawatts relative to leased megawatts, the credit quality of the tenant signing each lease, and the actual capex required per megawatt once ground is broken.
It also depends on the financing structure chosen to bridge the distance between today’s cash and tomorrow’s revenue, and on whether each company’s governance can support capital allocation at infrastructure scale.
Wall Street has already decided these companies are worth more as AI infrastructure than as Bitcoin miners.
What remains unsettled is whether investors are paying for AI cash flow that has not yet materialized, or for a construction pipeline that still needs tens of billions of dollars before it becomes AI revenue at all.
