Summary

  • CleanSpark discloses a signed 20-year triple-net lease for 175 MW of critical IT load at Sandersville, with two five-year tenant extension options. It expects deliveries to begin in the fourth quarter of 2027; that is not a disclosure of capacity already delivered or revenue already recognised.
  • The Texas portfolio is covered by an LOI and exclusivity arrangement involving 718 acres and up to 885 MW of secured and planned power capacity. The filing does not call it a lease, a definitive tenant commitment or contracted revenue.
  • At 30 June 2026, CleanSpark reported no AI and HPC-services revenue. Its expected fiscal-2028 mining decommissioning, financing, construction, equipment, approvals and power availability are still distinct gates.

One document contains three ledgers

The attraction of a large number is understandable. “Up to 885 MW” is easy to repeat and, placed beside the 175 MW Sandersville figure, looks like a single expansion programme. The company’s own disclosure does not support that compression.

Sandersville has a specifically identified contractual instrument: a 20-year triple-net infrastructure lease for 175 MW of critical IT load, entered on 10 July 2026. The tenant is described only as a leading global technology company. It holds two five-year extension options. These are precise commercial facts. They establish a lease relationship and a stated load figure. They do not disclose the tenant’s identity, its workload, a construction budget, an election of either extension or an operating date before delivery.

The second record is a Texas LOI and exclusivity arrangement. It concerns 718 acres and up to 885 MW of secured and planned power capacity. The difference in legal language is the analytical point, not a footnote. An LOI and exclusivity can frame a negotiation or reserve a lane for diligence. They are not, merely by being adjacent to a signed Sandersville lease, proof that Texas has its own definitive lease, committed tenant, energised load or rental stream.

The third record is physical conversion. CleanSpark says the current Sandersville site will continue bitcoin mining until lease commencement, then power will be diverted to the AI campus. It expects fully to decommission existing mining operations there during fiscal 2028. An expected decommissioning is neither a completed shutdown nor an accounting result. The filing does not quantify a resulting impairment or other accounting impact. The operating asset, the lease and the prospective Texas arrangement must each keep their own clock.

Critical IT load is a contract term, not a commissioning receipt

The filing says the tenant will lease 175 MW of critical IT load. That tells a reader what the lease is intended to provide. It should not be stretched into a claim that 175 MW is already energised, physically delivered, occupied by equipment or producing data-centre service revenue.

CleanSpark states that deliveries are expected to begin in the fourth quarter of 2027. It also describes milestones for financing, construction and delivery. The consequences matter: late completion can trigger rent abatements or termination. Those provisions are not proof of failure. They are proof that the lease’s commercial result depends on execution conditions that remain open.

The same filing names the conditions around those milestones: additional capital, likely project-based debt financing, specialised equipment, regulatory approvals, ongoing power availability and third-party development performance. Each represents an evidence gate with a different owner. A financier can decline or price debt. A supplier can miss a date. An approval can lag. Power can become unavailable. A developer can underperform. A signed lease changes incentives across that chain, but it does not eliminate the chain.

The relevant market question is therefore not whether the announced load is impressive. It is what receipt moves it from a contractual requirement to delivered capacity. A financing close, construction milestone, delivery notice and a later operating or revenue disclosure would answer distinct parts of that question. A repeated megawatt headline would not.

Texas is a separate option ledger

The Texas description combines three qualifiers: an LOI, exclusivity and capacity stated as secured and planned. No qualifier is decorative. “Up to” is a ceiling rather than a completion receipt. “Planned” identifies capacity that is not necessarily in service. The filing’s choice not to describe a Texas lease remains meaningful even though the same global technology company is said to have executed the arrangement.

Treating the 885 MW as a second leg of the Sandersville lease would create an unsupported aggregate. It would also hide what must happen next: a definitive agreement, terms that identify the parties’ binding obligations, a financing and development route, power availability evidence and the same construction-and-delivery evidence that Sandersville still needs. Until those arrive, Texas is strategic optionality with a disclosed negotiation boundary, not an operating portfolio.

This does not make the LOI irrelevant. Exclusivity can matter because it narrows whom the owner can negotiate with while a counterparty evaluates a site. But the value of such a right is conditional. It can expire, change, produce a definitive lease or not. The correct way to carry it into market analysis is as a separate state, rather than inflating it to match the legal certainty of Sandersville.

Revenue is the final ledger, and it is blank

The company reports that it had no AI and HPC-services revenue as of 30 June 2026 or 30 September 2025. That is unusually helpful negative evidence. It says neither signed lease language nor an LOI should be read backward as current service revenue.

The absence of revenue does not contradict the lease. It locates it. A development can be commercially important long before it is revenue-generating. Yet that period is also when the most consequential risk is unresolved: capital must be assembled; facilities and equipment must be delivered; approvals and power must hold; mining must be wound down in a managed sequence; and the tenant relationship must reach the milestones contemplated by the contract.

There is a capital-allocation consequence. The mining site may be a useful starting asset, but conversion is not costless simply because it reuses a location. Continued mining until lease commencement also means the old operating use has not yet ceded to the new one. The switch is a decision with timing, power-allocation and execution consequences, not an accomplished fact hidden inside an AI label.

What the market can test next

The next receipts should be kept in order. For Sandersville: financing, construction and delivery evidence; then evidence that critical IT load has become available; then service and revenue reporting. For Texas: a definitive agreement, if one is signed, followed by its own project and power evidence. For the conversion: a concrete decommissioning record and any disclosed accounting treatment, rather than an assumption that fiscal 2028 has already happened.

That sequence preserves the economic signal in CleanSpark’s disclosure. The company has disclosed one meaningful signed lease. It has also disclosed a distinct Texas negotiation position and a planned change in an existing mining operation. A market account that leaves those ledgers separate is not less ambitious. It is more useful: it shows where contract certainty ends and where delivery, operating proof and revenue still have to begin.

Sources