Stargate Lordstown: A named 2026 target needs delivered service evidence
FJI source-qualified: four-document gate met, including project-specific government evidence. Qualification evidence and decisions →
Project status: Groundbreaking disclosed September 2025; 2026 operating target unverified. Capacity scope: 1.5 GW combines Lordstown and Milam; Lordstown allocation undisclosed. How we verify evidence.
← Back to project indexCompare this project → · Track its catalysts → · Evidence guide →
- Location
- Lordstown, Ohio — representative locality pin
- Readiness
- Under construction
- Status
- Groundbreaking disclosed September 2025; 2026 operating target unverified
- Confirmed companies
- SoftBank; OpenAI
- Next milestone
- Dated 2026-target confirmation, site capacity allocation and commercial acceptance
- Reviewed
- October 6, 2026
Expanded analysis · Original disclosures linked below · CAD and USD preserved as labeled · Company statements attributed; interpretations labeled.
The one-minute investment takeaway
Our analysis: A named 2026 target needs delivered service evidence. Lordstown offers an Ohio development backed by major AI infrastructure participants, but portfolio capacity must be separated from campus capacity before valuation.
Readiness basis: Classification of evidence in the reviewed profile; strongest documented phase, not full-campus completion. Groundbreaking disclosed September 2025; 2026 operating target unverified Readiness source →
What the sources establish
- Named development. SoftBank’s September 24, 2025 release identifies Lordstown and Milam as SoftBank / OpenAI developments, with groundbreaking and operation expected the following year. The resulting 2026 target remains unverified. Primary source →
- Scope and participants. The same release places 1.5 GW across the two SoftBank-developed sites and separately describes Oracle-developed locations. Neither the combined capacity nor Oracle’s separate role is assigned entirely to Lordstown. Primary source →
Who receives the money?
Our analysis of disclosed roles. Revenue routes are business-model interpretations, not estimates of booked project revenue.
| Company / exposure | Economic role | What investors need to establish |
|---|---|---|
| SoftBank | Owner or operator of the disclosed infrastructure | Accepted useful capacity and net contribution after resource and renewal costs. |
| Named service and construction participants | Contract-specific delivery and recurring service roles | Actual contract, tariff or purchase terms; participation is not booked profit. |
| Equipment suppliers not established here | Potential procurement and renewal spending | Named awarded orders before attributing public-company project revenue. |
How financially meaningful is it?
Our analysis: The developer and computing customer have different economic exposures: delivering a powered facility can generate property or service contribution, while useful AI computing must cover facility charges, electricity, networking and hardware renewal. The disclosed partnership does not provide every ownership, lease or financing term, so an investor model should state which cash flows remain assumptions. SoftBank’s public stock provides corporate exposure to a broad group rather than a pure investment in Lordstown. The 1.5 GW combined figure cannot be multiplied by campus rent or allocated equally without evidence. An original 2026 target can be informative even late in the year, but it requires dated confirmation of commissioned facilities, accepted power and workloads. Oracle participation at other Stargate sites is not proof of an Oracle lease here.
Computing contribution equals customer or product benefit minus power, network, facility operation, financing and equipment renewal. Model long-lived buildings separately from shorter-lived computing assets.
Hypothetical illustration: 10 MW steady IT load at 1.2 PUE uses 105,120 MWh annually. A USD 10/MWh tariff increase adds USD 1.0512 million annual cost. Neither load, PUE nor tariff is disclosed here; this is a sensitivity example, not a site forecast.
Milestones that would change the assessment
| Milestone | Evidence to watch |
|---|---|
| Scope and capital | Reconcile campus and portfolio figures, accepted versus potential capacity, and property versus equipment funding. |
| Electrical and cooling acceptance | Dated 2026-target confirmation, site capacity allocation and commercial acceptance. Equipment delivery, construction completion and commissioning are separate steps. |
| Commercial productivity | Useful workloads or billable customer capacity, with operating costs and renewal obligations included. |
| Counterparty attribution | Named binding service and procurement terms before estimating participant earnings. |
What could break the investment case?
Our analysis: Portfolio capacity and commercial roles can be misallocated. Power, buildings and hardware may arrive on different schedules, while the original target may change without a public update.
What would change our view?
Our analysis: Dated acceptance, disciplined incremental capital and useful output at competitive resource cost would strengthen our assessment. Idle equipment, repeated delivery slippage or capital and renewal spending ahead of economic benefits would weaken it. We change the stage only on project-specific evidence, and would improve public-company financial attribution only with contract scope and cash-flow detail. A larger headline by itself would not justify a stronger return conclusion.
What remains unknown
Current accepted IT and facility MW, workload mix, utilization, site-attributable income, contract prices, tariffs, hardware renewal schedule and phase-specific capital allocation remain unestablished in the reviewed sources unless explicitly disclosed above. Targets remain expectations until confirmed. The map pin represents the locality rather than a verified individual building. Supplier, fund and corporate exposures must be modeled separately.
Project research for information, not a buy recommendation. Financial illustrations use assumptions, not disclosed site results or management guidance.
Scoped evidence update · October 17, 2026
In September 21, 2026 village testimony, company representatives described a modular R&D proof-of-concept, a before-year-end operating target and a foundation they said was in place. Those assertions are not independent inspection or commissioning evidence. Posted October 5 minutes record a 6–0 vote for Ordinance 57-2026, authorizing a cost-funding agreement on a non-emergency basis, effective 30 days after passage; site-plan requirements remained open. The signed agreement, current acceptance and complete corporate participant mapping remain unverified.
Source 1 · Source 2 · Source 3 · Source 4
Next checkpoint: October19 minutes approval/follow-up when available; full Ordinance57-2026 agreement; PoC acceptance against year-end target
Research reconciliation, not whole-project certification. FJI evidence · Documented company exposure
Company investor information & quarterly statements
Corporate-disclosure certification held: Official SkyVault entity/ownership and public/private financial-disclosure evidence; scope-confirmed OpenAI relationship
SoftBank Group
SoftBank developer named in original announcement; later village-recorded representatives describe a modular proof-of-concept. SkyVault, Crescent Dune and OpenAI scope/ownership require separate reconciliation.
Official investor & shareholder information · Quarterly financial statements: Q1FY2026 three months ended June30 2026
SkyVault / current PoC participant mapping
Representative-attributed PoC design/development role in September 21 village minutes
Disclosure verification held: Official legal entity, ownership and public/private financial-disclosure evidence remain unresolved; current OpenAI PoC relationship also unconfirmed.
Corporate disclosures provide company-wide context. They do not establish this project’s revenue, spending, profit, ownership share or operating capacity. Disclosure verification records →