Hyperscale Data Michigan AI Data Center Tops $70 Million, but $3 Billion Remains Conditional

Hyperscale Data has invested more than $70 million in its Michigan AI data center, while pushing its first contracted operations to November 2026. The spending signals real construction progress, but it does not make the widely cited $3 billion revenue figure guaranteed.

That larger figure assumes an unnamed customer expands from 20 megawatts to 52 megawatts and remains through every contract extension. The initial commitment presents a more grounded opportunity: 20 megawatts under a 10-year agreement, with two optional five-year extensions.

The distinction matters because Hyperscale Data is converting a Bitcoin mining site into infrastructure for a California-based neocloud provider. It must finish the deployment, finance remaining work, satisfy the contract, and manage local opposition before the projected revenue becomes operating cash flow.

The Michigan AI Data Center Has Entered Its Execution Phase

Hyperscale Data now has a signed customer, purchased equipment, and more than $70 million committed to the Michigan conversion.

The company disclosed the investment total on September 17, based on spending through September 15. It said the capital covered Alliance Cloud Services and development of the Dowagiac facility.

Alliance Cloud Services, or ACS, is the Hyperscale Data subsidiary operating the campus. The company says it has acquired a substantial amount of the equipment needed for the first contracted deployment.

That language indicates progress beyond a conceptual expansion plan. However, the company has not published a detailed equipment inventory, completion percentage, or remaining capital requirement.

The September update set November 2026 as the target for beginning operations under the master services agreement. That target represents a shift from the original contract schedule.

A June SEC filing targeted September 21 for the first 10-megawatt phase. It called for another 10 megawatts by the end of 2026.

The latest company update no longer describes a September start. Instead, management says operations should commence in November, when revenue and cash flow are expected to begin.

That is not necessarily a major delay for a complicated data center conversion. Still, it creates the first measurable test of management’s execution claims.

The contracted workload will occupy part of an existing industrial property in Dowagiac, Michigan. Hyperscale Data acquired the facility in 2022 after it had already been adapted for cryptocurrency mining.

The building covers approximately 617,000 square feet, according to industry reporting. The first customer deployment requires a retrofit of about 60,000 square feet.

Earlier disclosures estimated the initial 20-megawatt project would require between $100 million and $120 million. Consequently, the announced investment represents substantial progress, but not necessarily the entire required program.

Critical AI compute capacity refers to power delivered to computing equipment, rather than the facility’s total electrical draw. Cooling, power conversion, and other supporting systems consume additional energy.

That difference becomes important when evaluating both construction requirements and the campus’s advertised expansion potential. Delivering 20 megawatts of critical capacity requires much more than connecting servers to an existing mining site.

ACS must install and commission electrical infrastructure, cooling, networking, security, and customer-specific service areas. It must also show that converted infrastructure can meet uptime and performance requirements.

The customer paid an initial nonrecurring service charge and provided a security deposit when the agreement was signed. Those payments offer some contractual support, but they do not eliminate delivery risk.

The master services agreement also contains service credits, penalties, termination rights, and other customary protections. Revenue depends on ACS meeting its obligations, not merely spending capital at the site.

This is why the latest update matters. Hyperscale Data has moved from promoting available land and power toward proving that it can operate contracted AI infrastructure.

The $3 Billion Figure Is an Upside Case, Not Current Backlog

The most eye-catching revenue number combines customer expansion with a full 20-year relationship, making it a conditional scenario rather than booked revenue.

The existing agreement covers 20 megawatts of critical AI compute capacity. It begins with a 10-year term and includes two five-year extension options controlled by the customer.

If both extensions are exercised, Hyperscale Data says the initial capacity would generate more than approximately $1.2 billion across 20 years. Even that estimate remains subject to performance and continued customer participation.

The customer also holds a right of first offer for another 32 megawatts. A right of first offer allows the customer to negotiate for available capacity before Hyperscale Data broadly markets it elsewhere.

The right does not require the customer to take that capacity. It also does not establish that all required power, financing, construction, and regulatory approvals are already available.

If the customer adds all 32 megawatts within the specified period, total deployment would reach 52 megawatts. The company says maintaining that footprint through 20 years would push cumulative contract revenue above $3 billion.

This calculation combines several events that have not happened. The first 20 megawatts must enter service, the customer must expand, both parties must perform, and the customer must exercise both extensions.

A company progress update presents the same assumptions. Its language repeatedly describes the revenue as expected or potential.

Contract value also differs from annual revenue, profit, and cash available to shareholders. A multiyear total spreads payments across a long operating period while the provider continues funding equipment, maintenance, energy, and financing.

The distinction is especially important for AI infrastructure agreements. Large cumulative values often accompany long terms because data centers require substantial capital before service begins.

Hyperscale Data has not disclosed the customer’s identity. It describes the counterparty as a California-based neocloud provider offering managed computing and high-speed storage.

A neocloud is a cloud operator focused on GPU-intensive workloads, often serving AI developers that need computing capacity outside larger hyperscale platforms. Demand can be strong, but customer credit quality varies.

Without an identified customer, outside observers cannot independently assess its balance sheet, financing, GPU access, customer base, or ability to support a 20-year relationship.

The agreement’s upfront payment and deposit provide meaningful evidence of commitment. They are still much smaller than the cumulative revenue being discussed.

Investors and enterprise buyers should therefore separate three layers of value. The first is the signed 20-megawatt commitment. The second is the optional contract term. The third is an unexercised 32-megawatt expansion.

Only the first layer represents committed initial capacity. The other layers can become valuable, but each carries additional assumptions and time.

This framing does not make the contract insignificant. A successful first deployment would give Hyperscale Data an anchor customer and an operating reference for future sales.

It does mean that $3 billion should not be treated as money already secured. The figure describes the upper edge of one customer relationship under favorable conditions.

Bitcoin Infrastructure Is Giving Way to AI Capacity

The central bet is that power previously serving Bitcoin mining can produce steadier, longer-duration revenue through AI colocation.

Hyperscale Data previously operated approximately 28 megawatts of Bitcoin mining capacity at the Michigan campus. Management plans to reallocate power as the neocloud customer’s equipment comes online.

Bitcoin mining and AI computing both require dense electrical infrastructure, cooling, and reliable connectivity. That overlap makes former mining sites attractive candidates for AI conversion.

The workloads are not interchangeable, however. AI customers generally expect stricter service guarantees, network performance, physical security, and operational support than an owner-operated mining fleet needs.

A mining operator can tolerate some curtailment when energy costs rise or Bitcoin economics weaken. A colocation customer running training or inference workloads expects contracted capacity to remain available.

That changes the operating model. Hyperscale Data is moving from controlling its own compute workload toward supporting another company’s business-critical infrastructure.

The company’s original contract release estimated the first retrofit at $100 million to $120 million. It also described the transition as a progressive reallocation rather than an immediate shutdown of mining.

That gradual approach provides flexibility. Mining can continue using available power while ACS installs and commissions AI capacity.

It also creates coordination risk. Construction teams must modify the site while existing equipment and electrical systems remain active.

The economic rationale is clear. Bitcoin mining revenue depends on digital asset prices, network difficulty, machine efficiency, and energy costs.

Long-term colocation contracts can produce more predictable recurring payments. They also move part of the commercial risk toward customer retention and service delivery.

Peers have pursued the same transition. Core Scientific converted parts of its mining infrastructure for CoreWeave, while Applied Digital developed campuses around long-term AI tenants.

Those companies offer a useful reference, not a direct valuation template. Site design, tenant credit, financing, power access, and construction responsibility differ across contracts.

Hyperscale Data’s project also starts at a smaller scale. Its initial 20-megawatt commitment is enough to validate operations, but it does not yet place the campus beside the largest AI developments.

The project’s value rests partly on speed. Existing power infrastructure and an industrial building can shorten development compared with constructing an entirely new campus.

Yet existing infrastructure can also require extensive replacement. Equipment designed around mining racks may not support the density, redundancy, and cooling needs of newer GPU clusters.

The November target will therefore reveal more than whether one customer receives power. It will test whether Hyperscale Data’s conversion model produces a usable AI environment on schedule.

A successful deployment would create evidence for lenders, partners, and future tenants. A further delay would suggest that the inherited infrastructure offers less acceleration than management expects.

The Campus Ambition Is Larger Than the Current Contract

Hyperscale Data describes a 340-megawatt opportunity, but only 20 megawatts are contracted for the initial deployment.

Management says a full 52-megawatt deployment would use less than 20% of the Michigan facility’s potential capacity. It estimates approximately 270 megawatts would remain for other customers.

That arithmetic creates an expansive development story. The current customer could become an anchor tenant while ACS markets additional capacity to AI infrastructure buyers.

However, potential capacity is not the same as available capacity. Power must be secured through utility agreements, generation, transmission, substations, and site-level equipment.

The company explicitly conditions further development on financing, engineering studies, infrastructure availability, utility arrangements, and regulatory approvals. Those qualifications are central to the outlook.

AI data center operators are competing for transformers, switchgear, generators, cooling equipment, and skilled contractors. Long procurement cycles can slow projects even after customer demand appears.

Hyperscale Data must also decide how it wants to own the asset. Management says it is evaluating continued development, partnerships, additional customer deployments, a separation, an ACS initial public offering, or a sale.

Days before the latest update, the company established a minimum sale threshold of $750 million for the Michigan facility. Management also said it believes the asset’s value could reach or exceed $1.25 billion.

Those figures represent management’s position, not an announced acquisition offer. No buyer, transaction structure, financing commitment, or closing schedule has been disclosed.

The range of strategic alternatives introduces a second uncertainty around the campus. Hyperscale Data is building the project while considering whether it should remain the long-term owner.

A sale could transfer future construction requirements to a better-capitalized operator. It might also limit Hyperscale Data’s participation in the campus’s longer-term revenue.

A partnership could reduce funding pressure while preserving some ownership. An IPO or separation would require market support, audited financial information, and a credible independent operating plan.

Continued ownership offers the largest exposure to successful expansion. It also leaves Hyperscale Data responsible for financing an infrastructure program much larger than the current contract.

The company’s latest quarterly filing shows why capital structure matters. It reported $36.8 million in cash and cash equivalents at June 30.

The same filing listed approximately $201.7 million in current liabilities. That included approximately $99.3 million in current notes and related borrowing obligations.

Hyperscale Data recorded a $49.1 million net loss for the first six months of 2026. It used approximately $9.9 million of cash in operating activities during that period.

The company also raised capital through common stock sales and borrowing. Its outstanding Class A share count increased substantially during the first half.

These figures do not prove that Hyperscale Data cannot complete the campus. They show that financing choices will influence debt exposure, dilution, ownership, and construction pace.

Management’s near-term task is therefore narrower than developing all 340 megawatts. It must bring the first customer online and convert spending into recurring service revenue.

That operating milestone would strengthen the case for every strategic alternative. Without it, larger capacity and valuation claims remain mostly projections.

Financing and Local Opposition Pressure the Expansion Story

The investment demonstrates commitment, but the project still faces financial, operational, and community risks that spending alone cannot resolve.

The first risk is funding. The previously estimated cost for the initial deployment exceeds the $70 million invested through September 15.

Hyperscale Data has not specified how much of the remaining work is contractually committed, already financed, or still dependent on new capital.

The second risk is schedule performance. The initial SEC filing targeted September 21 for Phase 1, while the latest update points to November.

The company has not provided a detailed explanation for the changed target. Readers should treat November as a management objective until commissioned capacity begins serving the customer.

The third risk is customer concentration. The first AI contract depends on one unidentified neocloud provider.

An anchor tenant can support financing and construction. It can also create exposure when a significant portion of early revenue depends on one counterparty.

The fourth risk concerns expansion rights. The extra 32 megawatts support the $3 billion headline, but the customer has not exercised that right.

ACS must first make the initial deployment work. The customer will then evaluate its own demand, financing, hardware supply, and experience at the facility.

The fifth risk sits outside the contract. Dowagiac residents have complained about persistent noise from the property’s existing operations.

Residents filed a lawsuit alleging that the facility produced a high-pitched sound around the clock. The city also adopted an industrial noise ordinance and issued fines over alleged violations.

Hyperscale Data disputed the city’s measurement methods, according to local noise reporting. Management said it was acquiring nearby properties and pursuing noise-reduction measures.

The dispute predates the initial AI deployment, but it remains relevant. Cooling and electrical systems for dense compute can introduce additional sound sources.

Community opposition can affect permits, operating restrictions, litigation costs, and the pace of future construction. It can also make a technically viable expansion politically difficult.

Hyperscale Data says the 52-megawatt customer deployment would leave room for approximately 270 megawatts of future capacity. Realizing that potential requires more than internal land and power studies.

The company will need workable relationships with utilities, regulators, local officials, neighbors, customers, and capital providers. Weakness in any one area can constrain the broader plan.

The project’s strongest evidence remains its signed 20-megawatt agreement and actual capital spending. Its weakest claims involve capacity and revenue extending decades into the future.

A balanced assessment should hold both facts at once. Hyperscale Data is no longer describing an entirely speculative conversion, but it has not validated the full campus thesis.

What to Watch After the November Target

Three signals will determine whether the Hyperscale Data Michigan AI data center becomes an operating platform or remains an ambitious conversion project.

The first signal is service commencement. Hyperscale Data needs to confirm that the first contracted capacity is commissioned and producing recurring revenue.

A useful update would identify delivered megawatts, the actual start date, and whether the customer has accepted the service area. Cash collection would further strengthen the evidence.

If operations begin in November, the company will have cleared its most immediate execution hurdle. Another delay would weaken the argument that the site offers a rapid path from mining to AI.

The second signal is the customer’s response after deployment. Expansion to 52 megawatts is the critical condition behind the $3 billion revenue scenario.

An exercised expansion right would demonstrate that the customer sees enough demand and operating confidence to deepen its commitment. Silence would leave the larger figure hypothetical.

The terms matter as much as the announcement. Investors should look for delivery schedules, required capital, one-time customer contributions, and any changes to extension assumptions.

The third signal is a concrete financing or ownership decision. Hyperscale Data has named several possible paths, but none currently defines the campus’s future structure.

A funded partnership, binding sale agreement, or disclosed project financing package would clarify who pays for the next phase. It would also reveal how much economic interest Hyperscale Data retains.

Quarterly reporting should show whether AI colocation begins contributing meaningful revenue and cash flow. It should also separate those results from Bitcoin mining and the company’s other businesses.

Readers should resist using total company revenue as a substitute for data center performance. Hyperscale Data still owns operations spanning financial services, industrial activities, defense technology, and other sectors.

The cleanest measure will be revenue tied directly to ACS and the Michigan agreement. Capital expenditures, operating margins, and customer payments will help show whether that revenue creates economic value.

The $70 million investment establishes that Hyperscale Data has placed substantial capital behind its strategy. The next milestone is not another estimate of long-term capacity.

It is a working 20-megawatt service that meets customer requirements, produces cash, and operates within local constraints. That result would make the expansion story more credible.

Until then, the $3 billion figure should remain what the contract says it is: a potential outcome dependent on expansion, extensions, financing, and sustained execution.

Watch the November commissioning update, any 32-megawatt expansion notice, and the next financing disclosure. Together, those signals will show whether this conversion can support a durable AI infrastructure business.