CleanSpark AI Data Center Lease Reveals the High-Stakes Shift Beyond Bitcoin

CleanSpark has attached Meta to a $6.6 billion lease, turning a former Bitcoin mining campus into a test of its AI infrastructure ambitions. The CleanSpark AI data center lease covers 175 megawatts of critical computing capacity in Sandersville, Georgia. Initial rent is scheduled to begin in late 2027.

The agreement is much larger than a routine property contract. It asks investors to believe that CleanSpark can become a dependable data center developer while it still operates a volatile Bitcoin mining business. It also requires the company to deliver a technically demanding facility before most of the contracted revenue begins.

CleanSpark first announced the 20-year lease in July without identifying its customer. A September investor presentation revealed that the tenant is Anviran, a wholly owned Meta subsidiary. That disclosure strengthens the credit story, but it does not remove the construction, financing, or concentration risks.

The larger contest is no longer simply CleanSpark against other miners. It is Bitcoin mining economics against long-term infrastructure leasing. Riot Platforms, Core Scientific, and other mining operators are pursuing versions of the same transition.

The CleanSpark AI Data Center Lease Is Now a Meta Contract

The most important new fact is the tenant’s identity, not the headline contract value.

CleanSpark’s original lease disclosure described the customer only as a high-investment-grade global technology company. The agreement covers infrastructure at the company’s Sandersville campus and runs for an initial 20 years.

CleanSpark expects approximately $6.6 billion in contracted payments during that base term. Two five-year extension options would take the relationship beyond the initial period. Those options belong to the tenant, so their potential value is not guaranteed revenue.

The September investor presentation identifies Anviran as the tenant and Meta as its parent. Meta will guarantee rent and operating expenses, according to the presentation. The facility is intended to support Meta’s AI, data-processing, and communications requirements.

That identification materially changes how the agreement should be evaluated. A lease backed by Meta carries a different counterparty profile from one backed by an early-stage AI cloud provider. Meta has a global advertising business capable of funding large infrastructure commitments.

The contract uses a triple-net structure. Under this arrangement, the tenant carries specified operating costs in addition to rent. CleanSpark projects an operating margin near 100 percent before financing costs, depreciation, corporate expenses, and other excluded items.

That distinction matters. Net operating income, or NOI, measures property revenue after direct operating expenses. It does not equal net income or free cash flow. Interest, construction spending, taxes, and corporate costs still have real economic effects.

Annual rent carries a 3 percent escalator, according to CleanSpark’s presentation. The company projects average annual NOI of approximately $330 million across the initial lease term. Rent is expected to begin after the first network hall is completed.

The current target is November 30, 2027, with construction completion expected in March 2028. Those dates make the agreement an execution story rather than an immediate earnings transformation. CleanSpark must spend heavily before the lease produces a full year of revenue.

The project will provide 175 megawatts of critical IT load. Critical IT load measures the electricity delivered to servers and related computing equipment. It excludes some power consumed by cooling and other supporting systems.

The facility is therefore not a simple reuse of existing mining machines. Meta will need data halls, electrical redundancy, cooling, networking, security, and operating standards suited to high-density computing. CleanSpark is developing a different infrastructure product on a site previously associated with Bitcoin mining.

That conversion creates the central tension. The contract provides long-term visibility, but its value depends on CleanSpark delivering an asset that does not yet operate at contracted scale.

Why Bitcoin Miners Are Competing to Become AI Landlords

AI customers need power quickly, while Bitcoin miners already control land, grid connections, and energy infrastructure.

Securing enough electricity has become one of the hardest parts of building large AI systems. New transmission lines, substations, and utility interconnections can take years. Mining companies entered this market with sites already designed around continuous, high-volume power consumption.

That overlap makes their portfolios valuable. A miner may already control land, transformers, utility agreements, and fiber access. Those assets can shorten parts of the development process for a new data center.

However, a Bitcoin mine and an AI facility are not interchangeable. Bitcoin mining often prioritizes low energy costs and rapid equipment deployment. AI infrastructure places greater weight on reliability, cooling, network latency, redundancy, and tenant-specific technical designs.

The economics are also different. Bitcoin mining revenue changes with Bitcoin prices, network difficulty, transaction fees, and energy costs. A long lease can replace part of that exposure with contracted payments from a corporate tenant.

CleanSpark’s latest quarterly results show why that trade looks attractive. In its fiscal third-quarter report, the company reported quarterly revenue of $138 million. That was 30.5 percent below the comparable period.

CleanSpark also reported a quarterly net loss of $239.8 million. Changes in Bitcoin’s fair value contributed to the result. The accounting volatility illustrates how strongly a mining company’s reported performance can move with its digital assets.

Management has not said that Bitcoin mining will disappear. CleanSpark still controls a large mining fleet and describes Bitcoin as a core part of its infrastructure strategy. The Sandersville deal instead introduces another way to monetize access to electricity.

That diversification is not unique to CleanSpark. Core Scientific has been converting facilities toward high-density colocation, which provides powered space for customer computing systems. Riot Platforms has signed data center leases while continuing its mining operations.

The industry’s pivot has accelerated because AI developers value energized land differently from miners. A site can produce Bitcoin with specialized machines, or it can support long-duration rent from an AI customer. Management teams must compare both uses against conversion costs and delivery risks.

For investors, the shift changes the metrics that deserve attention. Hash rate and Bitcoin production remain relevant, but contracted megawatts, construction schedules, tenant credit, and financing terms become equally important.

For enterprise technology buyers, the transition also affects supply. New AI capacity increasingly depends on operators with roots outside traditional cloud computing. Buyers must evaluate both the tenant’s technology and the developer responsible for its physical infrastructure.

This is why the CleanSpark AI data center lease matters beyond one company. It shows that scarce grid access can move from cryptocurrency production to long-term AI infrastructure without changing ownership of the underlying site.

Meta’s Commitment Raises the Standard CleanSpark Must Meet

A strong tenant reduces credit risk while increasing the consequences of any delivery failure.

Meta’s involvement validates the commercial appeal of Sandersville. The company would not enter a long-term infrastructure agreement without extensive review of power availability, site design, construction planning, and counterparty capabilities.

Yet Meta’s name does not certify that the facility will arrive on time. CleanSpark remains responsible for coordinating development and satisfying contractual milestones. The project still faces permitting, equipment, utility, financing, and contractor risks.

The original lease announcement warned that missed milestones could lead to rent reductions or termination. It also identified electrical availability and construction financing as material uncertainties. Those warnings deserve equal weight with the projected contract value.

CleanSpark estimates a development cost of about $11.9 million for each megawatt of critical IT load. Across 175 megawatts, the total project requirement reaches roughly $2.08 billion before some financing-related reserves.

That figure explains why the company cannot treat the lease as a passive income stream. It must first assemble the capital and complete a complex construction program. The $6.6 billion headline represents payments spread across two decades, not cash received when the lease was signed.

CleanSpark says it has funded its anticipated equity contribution and prepaid long-lead equipment. Long-lead items are components that require substantial time between ordering and delivery. Procuring them early can reduce schedule risk, but it cannot eliminate construction delays.

The company is also pursuing project debt. A September financing announcement proposed $2.227 billion of senior secured notes due in 2031. The proceeds are intended to fund remaining construction, reimburse prior equity contributions, and establish debt-service reserves.

The notes would be secured by project assets and ownership interests in relevant subsidiaries. CleanSpark would also provide a completion guarantee. That guarantee requires the company to cover funding gaps if note proceeds do not complete the facility.

This structure can contain some risk within the project, but it does not make CleanSpark indifferent to overruns. A completion guarantee connects the parent company to delivery. Higher costs or delays can still consume capital that might otherwise support mining or additional data centers.

Meta’s guarantee addresses a different problem. It supports the tenant’s payment obligations after CleanSpark delivers the contracted infrastructure. It does not finance construction or ensure the developer completes the site.

Those two guarantees therefore sit on opposite sides of the transaction. Meta supports the lease payments. CleanSpark supports completion. The project succeeds only if both obligations function as planned.

The arrangement gives CleanSpark a credible customer, but also a demanding benchmark. Meta operates infrastructure at global scale and has detailed technical requirements. A developer entering this market cannot rely on mining experience alone.

CleanSpark must show that it can manage high-density cooling, resilient power distribution, network connectivity, and phased delivery. It must also coordinate those systems around Meta’s schedule without allowing its existing operations to distract management.

A signed lease is meaningful evidence of demand. It is not evidence that construction is finished, service levels are proven, or projected margins have reached CleanSpark’s financial statements.

The Real Contest Is Contracted Rent Versus Execution Risk

CleanSpark is exchanging direct Bitcoin exposure for a leveraged infrastructure build with delayed cash flow.

The company’s strategic reversal looks attractive because contracted revenue appears more predictable than mining income. However, predictability after delivery does not remove uncertainty before delivery.

CleanSpark plans to spend billions on an asset built for one major tenant. That creates concentration risk. Meta’s credit quality is strong, but the project’s economics depend heavily on one customer, one campus, and a specific technical design.

The base lease lasts 20 years and includes annual escalators. Long duration can support financing because future rent helps lenders model debt service. It can also limit flexibility if technology requirements change faster than the agreement.

AI hardware is evolving quickly. New accelerators can require different rack densities, cooling systems, and network architectures. CleanSpark must build infrastructure that meets current specifications while retaining enough flexibility for future equipment generations.

The triple-net format transfers many property operating expenses to the tenant. That supports CleanSpark’s projected NOI margin. It does not transfer all capital costs, debt expenses, or obligations created by the completion guarantee.

CleanSpark’s presentation uses an illustrative 7.5 percent coupon in its project model. Actual financing conditions ultimately determine how much contracted rent remains after debt service. A higher borrowing cost would reduce the cash available to the developer.

Timing creates another risk. The first rent is scheduled for late 2027, while capital must be committed much earlier. Any delay extends the period during which the project consumes cash without producing its planned lease income.

There is also a difference between contract value and economic value. Adding every scheduled payment produces the $6.6 billion figure. Investors still need to discount future payments, subtract financing costs, account for construction spending, and consider delivery probabilities.

The extension options deserve similar discipline. CleanSpark has discussed a larger potential value if Meta uses both five-year extensions. Those payments should not be treated as committed because Meta controls the decision.

The Texas opportunity remains less certain. The same customer entered a letter of intent and exclusivity arrangement covering CleanSpark’s Texas portfolio. That portfolio represents up to 885 megawatts across the Sealy and Brazoria campuses.

A letter of intent is not a signed lease. Exclusivity can prevent CleanSpark from negotiating with other parties during a specified period, but it does not guarantee that Meta will contract for all available capacity.

The Texas discussions still matter because they point toward a relationship larger than Sandersville. They also increase the importance of the Georgia project. Successful delivery could support additional contracts, while a significant failure could damage CleanSpark’s credibility.

The most useful way to read the CleanSpark AI data center lease is therefore as an exchange of risks. CleanSpark reduces its dependence on Bitcoin prices for the contracted site. In return, it accepts development, financing, tenant concentration, and long-duration execution risk.

That trade can create a more durable business. It can also expose a mining operator to responsibilities that traditional miners never had to manage.

Riot and Core Scientific Show This Is an Industry Race

CleanSpark is joining an established competition to convert scarce electrical capacity into AI infrastructure contracts.

Riot Platforms offers the clearest direct comparison. Riot has kept its mining business while marketing its land and power portfolio to data center customers.

In January 2026, Riot announced a lease with AMD for an initial 25 megawatts at its Rockdale, Texas site. The AMD agreement carries a 10-year initial term and expansion mechanisms covering additional capacity.

Riot expected the first deployment to use existing infrastructure at Rockdale. Its disclosed retrofit cost was lower per megawatt than CleanSpark’s projected Sandersville development cost. However, the projects differ in size, design, lease structure, and construction scope.

The comparison still reveals a common strategy. Both companies use power portfolios assembled for Bitcoin mining to win contracts from major technology customers. Both retain exposure to mining while building a second infrastructure business.

Core Scientific represents a more advanced conversion. The company has said it intends to repurpose most of its facilities for AI-related workloads and high-density colocation. It expects those operations to become a much larger source of revenue.

Its experience also provides a warning against treating every AI transaction as certain. Core Scientific shareholders rejected a proposed acquisition by CoreWeave in 2025. The companies retained their operating relationship, but the failed transaction showed that strategic logic does not settle valuation questions.

A Core Scientific filing also lists familiar risks. Those include limited power availability, capital requirements, customer concentration, regulatory changes, and the difficulty of converting mining sites.

These examples pressure miners that have not secured major tenants. Once competitors show that grid-connected campuses can attract technology companies, shareholders can ask why other portfolios remain devoted entirely to Bitcoin.

They also pressure traditional data center developers. Mining companies bring large power positions and experience operating energy-intensive facilities. Established developers counter with deeper construction, leasing, and uptime records.

The competitive divide is not simply old data centers against new entrants. It is a race among companies that control power, capital, construction expertise, and credible routes to customer delivery.

CleanSpark’s Meta agreement puts it in that race with a substantial contracted project. It does not automatically place the company ahead. Riot has already delivered capacity to AMD, while Core Scientific has longer experience with high-density hosting.

The companies also differ in how they fund development. Some sell Bitcoin, some issue corporate securities, and others arrange project-level debt. Each method changes shareholder exposure and the amount of risk retained by the parent company.

Customers should examine those financing choices because undercapitalized development can become an operational problem. A tenant depends on the developer completing the site, maintaining infrastructure, and responding to failures throughout the lease.

Investors should resist valuing every contracted dollar equally. A delivered megawatt with operating history deserves different treatment from planned capacity. A signed lease deserves different treatment from an exclusivity letter.

This industry race will ultimately be decided by commissioning, uptime, and cash flow. Announced megawatts establish a pipeline, but delivered megawatts establish a business.

Three Signals Will Determine Whether the Pivot Works

The next evidence must come from financing, construction, and Texas contracting, in that order.

The first signal is the final Sandersville financing package. CleanSpark proposed $2.227 billion in secured notes, but the announcement made completion subject to market conditions. Investors need the final principal, interest cost, covenants, and reserve requirements.

Those terms will show how lenders price CleanSpark’s execution risk and Meta-backed rent. Favorable financing would strengthen the case that the lease can support itself after construction. Expensive or restrictive debt would weaken projected cash returns.

The second signal is progress toward the November 2027 rent date. Equipment procurement, utility work, construction milestones, and network-hall completion will matter more than broad statements about AI demand.

CleanSpark says long-lead equipment has been ordered and prepaid. The next step is showing that site work remains aligned with the delivery schedule. Any revision to the completion date, budget, or phased capacity would deserve close attention.

The third signal is whether the Texas exclusivity arrangement becomes a binding lease. Meta’s interest covers up to 885 megawatts, which is several times the Sandersville capacity. A signed Texas agreement would make this a portfolio relationship rather than a single-site transaction.

Failure to convert the letter of intent would not invalidate Sandersville. It would, however, weaken the idea that one successful negotiation can unlock CleanSpark’s broader power portfolio.

Readers should also watch how management reports the transition. Separating data center development results from mining performance would make the economics easier to assess. Clear reporting should distinguish signed leases, nonbinding discussions, construction spending, and operating capacity.

The same discipline applies across the mining sector. Contract value is useful, but it cannot replace information about delivery, leverage, customer obligations, and recurring cash generation.

For developers and enterprise buyers, the story highlights a broader infrastructure dependency. AI services do not begin with a model endpoint. They begin with land, electrical interconnections, cooling equipment, financing, construction labor, and long-term operating agreements.

Teams evaluating AI suppliers should document those dependencies alongside software risks. A searchable knowledge base can help procurement, engineering, and finance teams preserve contract assumptions and track changes across technical documents.

The CleanSpark AI data center lease has passed its first major credibility test by naming Meta and outlining a financing path. Its harder test is still ahead. CleanSpark must deliver 175 megawatts of production-ready capacity before the headline value becomes operating performance.

Watch the debt terms, the construction schedule, and the Texas negotiations. Together, those signals will show whether CleanSpark is becoming an AI infrastructure landlord or simply financing an expensive transition narrative.