Soluna Project Kati 2 Moves Forward at $471.8M, but Power Is Only the First Test

Soluna Project Kati 2 has moved closer to construction with a $471.8 million filing for a 112-megawatt AI data center in South Texas. The new record gives physical scale and a proposed schedule to a campus that Soluna Holdings and Metrobloks first presented as a renewable-powered computing project.

The filing sharpens the story, but it does not settle it. Project Kati 2 still needs substantial outside financing, binding customer commitments, power approvals, and a workable operating plan. Its developers have secured a promising location near wind generation, yet available energy alone does not produce a functioning AI campus.

That gap between a power-rich site and a financeable data center is the project’s central tension. Texas is receiving an extraordinary volume of large-load requests, while ERCOT is tightening how it evaluates projects. Soluna and Metrobloks must now prove that Kati 2 is more advanced than the speculative capacity surrounding it.

Soluna Project Kati 2 Now Has a Building Plan

The latest filing turns Project Kati 2 from a broad campus proposal into a defined construction project, although a filing is not proof of completion.

The planned facility would cover 471,884 square feet near FM 1420 in the Santa Monica area of Willacy County. It would be a single-story data center with an estimated construction cost of $471.8 million.

The project filing details also identify approximately 112 megawatts of critical information technology load. Critical IT load measures the electricity delivered to computing equipment, rather than the facility’s entire electrical demand.

That distinction matters. Servers, networking systems, cooling equipment, power conversion, lighting, and other infrastructure all consume energy. A 112-megawatt critical IT load therefore implies a larger total facility requirement.

The filing scheduled construction to begin on October 8, 2026, with completion targeted for October 8, 2027. Those dates describe the registered plan. They do not independently confirm that full construction mobilization occurred on schedule.

The 112-megawatt figure is also higher than the initial development described in January. Soluna and Metrobloks originally outlined a first phase exceeding 100 megawatts, followed by a longer expansion path.

That increase is modest in percentage terms, but meaningful in infrastructure terms. Twelve additional megawatts can represent thousands of high-performance processors, depending on hardware, utilization, and supporting systems. It also adds pressure to power delivery and cooling plans.

Soluna Project Kati 2 is separate from Kati 1, which forms another part of the company’s Willacy County footprint. Soluna broke ground on the broader Project Kati development in September 2025 and described it as a wind-powered campus.

Kati 1 was organized around an earlier 166-megawatt plan divided into two 83-megawatt phases. Kati 2 has since emerged as a separate joint venture focused on AI and high-performance computing, or HPC.

The new facility is not simply a larger version of a cryptocurrency mining site. AI infrastructure requires different network, cooling, availability, and equipment-density assumptions. Training clusters also depend on synchronized servers and high-bandwidth connections that make interruptions more complicated.

The filing therefore creates a clearer test. Soluna and Metrobloks have attached a construction budget, building footprint, load estimate, and proposed deadline to their plan. Those details make progress easier to measure, but they also expose any later delays.

The most important immediate change is not the headline investment. It is the movement from a partnership narrative toward an identifiable building program. Investors, customers, utilities, and local officials can now compare actual milestones against the October 2027 target.

Why South Texas Has Become Part of the AI Infrastructure Race

Project Kati 2 is a bet that access to generation will matter more than proximity to an established data center market.

Northern Virginia, Dallas, Phoenix, and other major data center regions offer mature fiber, contractors, suppliers, and customers. They also face competition for electricity, land, transmission access, and equipment.

Willacy County offers a different starting point. It sits near substantial wind generation and outside the most crowded Texas data center clusters. Soluna’s strategy is to place flexible computing demand close to renewable plants that sometimes produce more electricity than the grid can efficiently absorb.

Developers call this a behind-the-meter model when electricity moves from nearby generation to the customer through infrastructure positioned before the conventional retail meter. The exact commercial and grid structure varies, but the objective is consistent: reduce dependence on distant transmission capacity.

Soluna has spent years promoting what it calls renewable computing. Its facilities can adjust consumption in response to energy availability, grid conditions, and market prices. That model began with Bitcoin workloads, which can pause more easily than many conventional data center applications.

AI changes the requirements. Some machine-learning workloads can be scheduled around periods of available energy, especially batch processing that does not serve live users. Large training jobs, however, can lose time and efficiency when thousands of processors stop unexpectedly.

Inference presents another challenge. Inference is the process of using a trained AI model to respond to requests. Customer-facing inference services often require low latency and consistent availability, leaving less room for interruption.

Kati 2 will therefore need more than inexpensive wind energy. It needs a power architecture that can support dense computing equipment through changing grid and generation conditions. It also needs cooling, backup systems, and operating procedures suited to customer service agreements.

Soluna contributes the site position, power work, electrical equipment, and development experience. Metrobloks brings data center design, leasing, operations, and customer engagement capabilities. The companies described that division of labor in their development announcement.

The partnership reflects a wider change in the AI infrastructure market. Securing processors remains difficult, but securing power has become equally important. A developer with land but no credible electrical path cannot promise a delivery date.

Metrobloks CEO Ernest Popescu summarized that pressure in January: “They need power and capacity now, not years from now.” The statement is promotional, but it captures the commercial problem behind the project.

South Texas also adds geographic diversity. Many announced Texas data centers concentrate around Dallas, Central Texas, or energy-producing areas farther west. A Willacy County campus creates another deployment option for customers serving Texas, Mexico, and Latin American markets.

Metrobloks says the site is within 40 miles of McAllen and near long-haul and cross-border fiber routes. Those claims will matter only if customers can obtain the network diversity, latency, and carrier capacity their workloads require.

The location also changes the project’s risk profile. A less congested market can offer space and power, but it may have fewer specialized contractors and established data center suppliers. Construction teams must move equipment, materials, and technical labor to the site at the required scale.

Project Kati 2 is consequently testing a broader proposition. AI capacity does not always need to sit inside a traditional technology hub, provided power and connectivity can support the workload.

That proposition pressures incumbent markets facing long interconnection timelines. It also pressures Soluna and Metrobloks to demonstrate that an alternative location can match the reliability customers expect from established campuses.

Available Wind Power Does Not Guarantee an AI Tenant

The project’s strongest advantage is its energy position, while its biggest weakness is the distance between that position and a contracted, financed campus.

Soluna and Metrobloks announced in January that a potential neocloud tenant had signed a non-binding letter of intent. A neocloud is a specialized cloud provider that rents GPU computing capacity, usually for AI training and inference.

A non-binding letter can establish interest and guide technical planning. It is not the same as a signed lease, committed revenue, or a creditworthy guarantee supporting construction debt.

That distinction becomes important for a project with a stated construction cost approaching half a billion dollars. Lenders and equity investors typically want evidence that customers will occupy the capacity and pay under enforceable agreements.

Soluna’s regulatory disclosures provide a more cautious view than the promotional announcement. The company’s joint venture filing says Kati 2 requires substantial third-party financing that had not been obtained as of June 30, 2026.

The filing also shows how the partnership is structured. Soluna’s subsidiary holds the voting interests and manages the joint venture. The Metrobloks affiliate holds profits interests, contributed no initial capital, and has limited protective rights.

Soluna reported a funded contribution of about $6.7 million when the arrangement was established. That included land, earlier development costs, and cash. It also committed additional capital for another property acquisition and certain operating expenses.

Those contributions advance development, but they are far below the filed construction estimate. The difference must come from future financing arrangements, customer-backed capital, or another project-level structure.

The joint venture also reported no third-party debt outstanding as of June 30. Its undeveloped land and capitalized development work did not generate revenue, according to Soluna’s disclosure.

This does not mean Kati 2 lacks a viable financing path. Large infrastructure projects often progress through land control, design, permitting, customer negotiations, and financing in parallel. However, each step remains conditional until contracts close.

The construction filing gives potential financiers a more specific asset to evaluate. It does not remove their questions about tenant credit, equipment procurement, power terms, construction risk, or operating performance.

AI data centers carry another complication. Their customers can be young companies whose demand grows quickly but whose long-term credit history remains limited. A neocloud may need substantial capital for GPUs while also committing to a long facility lease.

Hyperscale cloud companies bring stronger balance sheets, but they can demand custom designs, strict delivery guarantees, and favorable commercial terms. They may also control several competing site options.

Soluna Project Kati 2 therefore faces a sequencing challenge. The developers need enough design certainty to win a tenant. They also need enough customer certainty to finance and finalize the design.

Power does not solve this loop by itself. It gives the project a valuable input and a persuasive sales argument. The eventual tenant must still accept the location, reliability design, network plan, sustainability claims, and delivery schedule.

Kati 2 also competes with a growing collection of Texas proposals. Some are backed by large cloud operators, established data center developers, energy companies, or vertically integrated AI infrastructure providers.

Not every announced megawatt will be built. Projects with confirmed land, credible power, binding tenants, financing, and ordered equipment will separate from projects that remain primarily in development queues.

The South Texas campus has progressed beyond a concept because its developers formed a joint venture, contributed property, and filed a detailed construction plan. Yet it has not crossed the line into a fully financed, fully contracted asset.

That is the most important counterweight to the headline. The $471.8 million estimate describes the anticipated construction scope. It does not show that the full amount has been committed.

Texas Grid Rules Put the 112 MW Plan Under Review

Kati 2’s location near renewable generation helps its case, but Texas no longer treats a large-load request as routine.

ERCOT is managing a wave of proposed electricity demand that far exceeds the state’s current system peak. In June 2026, the grid operator said it was tracking more than 438,000 megawatts of large-load requests.

ERCOT attributed nearly 89 percent of that requested capacity to data centers. Many proposals overlap, lack final customers, or will never reach operation. Even after discounting speculative entries, the volume changes how the grid evaluates new demand.

Texas responded with a batch process for projects of at least 75 megawatts. Instead of studying every large load in isolation, ERCOT can evaluate groups together and identify how simultaneous connections affect reliability and transmission.

The Batch Zero process is designed to determine which loads can connect, where the grid can support them, and what upgrades are necessary. At 112 megawatts of critical IT load, Kati 2 is well above the threshold.

Its actual grid interaction may be more complicated than the critical IT figure suggests. Total facility demand is higher, while behind-the-meter generation and operating flexibility can reduce or reshape the amount requested from the grid.

Soluna says the campus will combine renewable power with ERCOT access. That dual arrangement is commercially attractive because it can offer renewable energy without relying entirely on intermittent wind output.

However, dual access creates technical and regulatory questions. ERCOT and the transmission provider need credible models showing when the facility will draw from the grid, how quickly demand can change, and how it behaves during disturbances.

Texas Senate Bill 6 added further requirements for large customers. The law created new interconnection protections, cost allocation measures, and demand-management obligations following concerns about rapid load growth.

Certain new large loads must support curtailment during emergency conditions. Curtailment means reducing electricity use when the grid faces a shortage or reliability event. For flexible computing operators, this can create both an obligation and a commercial opportunity.

Soluna already has experience adjusting Bitcoin workloads. AI clusters may not offer the same flexibility without careful workload design. A training job spanning thousands of GPUs cannot always stop without lost progress or recovery time.

The project will need to distinguish among workloads. Batch jobs can sometimes move to periods with better wind output. Live inference, networking, cooling, and storage systems require greater continuity.

Backup generation, batteries, redundant feeds, and workload controls can bridge some interruptions. Each adds expense, maintenance obligations, emissions considerations, or operating complexity.

Water is another open issue. The public filing identifies the building size and construction estimate but does not establish the facility’s expected water consumption or final cooling design.

Metrobloks describes its broader designs as using closed-loop chilled-water systems with low water consumption. That company-level description does not independently verify Kati 2’s eventual equipment or annual water demand.

The issue is especially important in South Texas, where communities must balance industrial growth with agricultural, municipal, and environmental needs. Different cooling architectures can produce very different water and electricity profiles.

In August, Texas directed ERCOT and the Public Utility Commission to gather more information about large data centers. The resulting community-impact review asks developers about water sources, consumption, cooling, grid dependence, on-site generation, tax benefits, noise, light, traffic, and ownership.

Developers of data centers at or above 25 megawatts were told to respond by October 12, 2026. ERCOT plans to publish an analysis by December 10.

That process will create a near-term credibility test for Kati 2 and its peers. Developers can no longer rely solely on a headline capacity number. State officials want evidence about how proposed facilities affect communities and the grid.

Project Kati 2 may benefit from being co-located with wind generation and designed for flexible computing. It still must demonstrate that those advantages work under the new review framework.

The developers should not overstate the environmental case before the final operating design is visible. Renewable proximity does not eliminate backup emissions, embodied construction impacts, transmission needs, or water use.

Likewise, critics should not assume a 112-megawatt facility will draw its maximum grid demand continuously. Load behavior depends on the customer, workload schedule, power contracts, cooling system, and available on-site resources.

The decisive information will come from specific agreements and engineering documents. Until then, the project’s renewable-computing promise remains credible but unproven at AI scale.

Three Signals Will Show Whether Kati 2 Is Real

The next phase will be decided by a tenant, financing, and regulatory clearance, not by another expansion announcement.

The first signal is a binding customer agreement. Soluna and Metrobloks need to convert reported tenant interest into a lease, capacity reservation, or comparable commitment.

The quality of that agreement matters as much as its existence. Investors will examine the customer’s credit, contract duration, power obligations, ramp schedule, and remedies for delayed delivery.

A binding agreement covering a meaningful portion of the initial capacity would strengthen the project’s financing case. Another non-binding expression of interest would add less certainty.

The customer’s intended workload will also reveal how Kati 2 plans to use energy. A training-focused tenant may tolerate more scheduling flexibility than an inference provider promising constant response times.

The second signal is a financing package aligned with the $471.8 million construction plan. Soluna’s June disclosure made clear that substantial third-party capital was still required.

A construction loan, infrastructure investment, development partner, or customer-supported funding structure would move the campus into a different risk category. It would show that outside parties completed meaningful diligence and accepted the project’s commercial assumptions.

The financing terms will be instructive. Expensive capital, extensive guarantees, or unusually high return requirements would suggest that investors still see major execution risk.

Equipment commitments matter here too. AI data centers need transformers, switchgear, cooling systems, backup resources, network equipment, and specialized construction materials. Long procurement periods can undermine an otherwise credible deadline.

The third signal is progress through ERCOT and the state’s impact review. Kati 2 needs an approved electrical path that matches its proposed 112-megawatt computing load and broader facility demand.

The December state report should clarify how regulators view water, community effects, on-site power, and the credibility of Texas data center requests. Project-specific disclosures may remain limited, but the review can still change requirements and timelines.

A clear interconnection position would support the October 2027 completion target. Required transmission upgrades, unresolved studies, or new restrictions would weaken it.

Construction activity provides a useful secondary check across all three signals. Site preparation, major equipment delivery, structural progress, and utility work are harder to replicate than announcements.

Readers should treat the filed dates as a benchmark. If visible work and financing milestones align with that schedule, the development case strengthens. If milestones repeatedly move, the headline estimate becomes less informative.

The broader stakes reach beyond one company. Kati 2 is testing whether renewable-rich areas can host dense AI infrastructure without copying the standard hyperscale campus model.

Success would give AI providers another route to capacity and give wind projects a large local customer for energy that transmission constraints might otherwise limit. It could also draw more infrastructure investment toward the Rio Grande Valley.

Failure would expose the limits of treating electricity access as the decisive advantage. Power is essential, but data centers also depend on capital, customers, networks, cooling, construction execution, and public acceptance.

Developers and enterprise buyers evaluating this market should record each commitment separately instead of treating every announced megawatt as equivalent. A searchable engineering knowledge base can help teams connect filings, contracts, grid decisions, and construction changes over time.

Soluna Project Kati 2 now has enough specificity to deserve attention. It has a named joint venture, a defined building, a 112-megawatt target, a proposed schedule, and a large construction estimate.

What it does not yet have publicly is equally important: completed third-party financing, a disclosed binding tenant, confirmed final cooling data, and a fully demonstrated interconnection path.

The next announcement that matters should close one of those gaps. Will Kati 2 secure a customer and capital before its construction schedule begins to slip, or will it remain another attractive power site waiting for an AI business case?