CloudHQ Querétaro: what is earlier usable capacity actually worth?

01 — Investment question and decision thesis

The commercial question is not whether a very large data-center development has been announced. It is whether a defined increment of electrically accepted, connected, commissioned and contracted capacity can produce distributable operating contribution earlier, and whether the incremental expenditure needed to change the critical path has positive risk-adjusted value. An announcement, a building shell, substation voltage, line rating, available power, IT load, customer commitment and invoiceable capacity describe different stages of value creation.

For LARGITAS, this creates two distinct potential assignments: a developer-side commissioning and capital-sequencing decision, and a supplier-side opportunity-qualification decision. The former should be evaluated with incremental discounted cash flow; the latter with addressable scope, bid probability, avoidable costs, timing and counterparty exposure. Neither is substantiated by multiplying announced megawatts by a guessed rental tariff.

02 — Chronology: what changed and when

• September 2025: CloudHQ's project presentation describes six planned data centers, up to 600 MW of critical computing load, initial power arrangements for the first 200 MW of a 900 MW campus electrical requirement, more than US$250 million of estimated predevelopment capital, US$4.8 billion of planned construction, a private campus substation up to 900 MW, and a 2 GW switching station supporting the wider grid. The 2027 dates are plans, not operational certifications. [S1]
• September 25, 2025: Mexico's federal government announces US$4.8 billion for six facilities on 52 hectares, and projected construction and permanent jobs. This is a public announcement rather than audited capital deployed. [S2]
• The corporate QRO campus webpage, accessed October 23, 2026, displays 360 MW of available/total critical IT load and describes six buildings of 48 MW IT each (288 MW arithmetically). It also gives 254,400 square meters without proving that figure represents the same land parcel as the government's 52 hectares. [S3]
• September 10, 2026: El Economista attributes to a CloudHQ executive a completed first 400 kV substation, 36% progress on another, roughly two years of land-related preparation, roughly three years linked to energy development and discussion of up to 920 MW. These are attributed remarks, not independent certificates of energization. [S4]
• October 5–13, 2026: local reporting, quoting an updated company/state meeting, describes a proposed increase from US$4.8 billion to nearly US$6 billion and nearly 1,000 MW of energy capacity. This is a later project-scope report. It cannot be treated as cumulative cash invested or as a completed capacity increase. [S5]

Commercial-contract status control: The September 2025 company presentation says that, following initial power arrangements, CloudHQ intended to secure a long-term lease with a leading global technology company and proceed with construction. That statement is evidence of a planned contractual milestone at the document date, not evidence that no lease has since been signed. A tenant-specific billing date and its actual status must be independently established before treating additional usable MW as cash-generating capacity. [S1]

03 — Capacity reconciliation: the foundational control

MW is active power, MVA is apparent power and kV is voltage. The conversion between MW and MVA needs power factor; the relation between facility electrical draw and IT load requires an explicit power-usage-effectiveness (PUE) definition, auxiliary consumption and relevant operating window. A simple ratio such as 900/600 = 1.5 is not proof of PUE because both numerators and denominators may describe different development phases or infrastructure boundaries. Actual net deliverability also depends on accepted interconnection, transformer ratings, redundancy, grid availability, customer provisioning and operating permits.

An additional primary-source terminology issue exists: the September 25, 2025 federal conference transcript verbally refers to roughly 900 MW of “critical computing load,” whereas the branded presentation distinguishes up to 600 MW of computing load from 900 MW of campus electrical load. This difference should be flagged as a possible spoken-definition or scope inconsistency, not silently resolved or alleged to be misconduct. [S8]

04 — Land, site footprint and construction status

The presentation and federal release describe 52 hectares. The corporate webpage lists 254,400 square meters, approximately 25.44 hectares, but its label is embedded in a campus-property description with a square-foot equivalent. That could be building area, a distinct parcel, phased floor area or another measurement. It must not be described as a verified land-boundary contradiction without survey documents. The correct diligence request is an indexed site master plan, property title schedule, cadastral polygons, building area program and phase-by-phase permits.

Public claims about substations being 'complete' have similarly narrow meanings. Civil completion, installed equipment, commissioning tests, utility acceptance, first energization, load-testing, firm power quality and redundant operation are separate milestones. The cash-flow model should begin only after the contractually defined service-availability or invoicing milestone, not merely after civil completion.

05 — Sector and grid context

Proyectos México reports 2025 electricity generation of 353,820 GWh, a 52,945 MW peak in the National Interconnected System, 111,195 km of transmission and 908,195 km of distribution, with approximately 49.8 million users. The national totals establish system scale, not spare capacity at CloudHQ's connection node. [S6]

Mexico's published electricity planning materials describe a Querétaro supply project, P24-OC3, and timing, transformation and transmission constraints. Secondary reporting attributes to the plan 675 MVA of transformation and 51.2 circuit-km. These figures should not be asserted as dedicated CloudHQ power without interconnection and ownership documents. [S7]

Regional electricity generation sufficiency is not sufficient evidence of substation-level deliverability. A data center can be delayed by right-of-way, utility interface, transformer supply, protection studies, fiber route diversity, water/cooling permits, traffic or tenant contract execution even when generation exists elsewhere.

06 — Decision unit and boundary of the model

Use ONE company-described 48 MW IT-load building as an analytical unit, strictly as a scale reference. Its modeled 65% billed/occupied utilization, US$70 per billed kW-month contribution margin, 120-month operating life, 12% effective annual discount rate and three-month service-start difference are assumptions, not CloudHQ operational data. The modeled US$70 is contribution after modeled variable costs, NOT gross rent, EBITDA, net profit or actual CloudHQ pricing. The 120-month economic life is held constant for the two alternatives to isolate timing rather than delete three months of operations.

The result is an incremental operating cash-flow timing experiment. It deliberately excludes construction draw timing, inflation, taxes, debt service, terminal value, vacancy ramp shape, tenant rebates, penalties, decommissioning and possible contract-end-date effects. In a real mandate these belong in a linked, project-specific financial model.

07 — Exact mathematics: three-month change in service date

• IT capacity reference: K = 48 × 1,000 = 48,000 kW.
• Hypothetical paid proportion: u = 0.65. Contribution margin: c = US$70 per billed kW-month.
• Monthly incremental operating contribution: C = K × u × c = US$2,184,000.
• Monthly discount factor: i = (1 + 0.12)^(1/12) − 1 = approximately 0.9488793%.
• On-time annuity value: PV0 = Σ(t=1..120) C/(1+i)^t = approximately US$156,058,886.
• Same 120 cash flows beginning three months later: PV3 = PV0/(1+i)^3.
• Value of advancing this hypothetical schedule: ΔPV = PV0 − PV3 = approximately US$4,359,439.
• Assuming 40% incremental probability the proposed intervention succeeds and US$750,000 paid regardless: expected incremental NPV = 0.40 × ΔPV − 750,000 = approximately US$993,776.
• Success probability needed to break even: p* = 750,000/4,359,439 = 17.20%.

The value is mathematically attributable to the timing of contribution, not automatic 'revenue recovered.' The intervention cannot produce these economics unless it actually improves the binding completion path and the capacity is paid for. If the contract has a fixed expiry rather than a fixed operating tenure, the delayed case must separately remove terminal cash flows; this model intentionally does not.

08 — Sensitivity and loss-producing alternatives

At only one month of actual schedule acceleration, the base contribution case gives approximately US$1.467m in timing value, and the same 40%-success/US$750k intervention has negative expected NPV of approximately US$163k. Six months would generate more timing value, but cannot be assumed unless the critical-path analysis supports six months. The model therefore rejects expensive schedule remedies when expected benefits do not justify them.

In project-specific work, build a three-way scenario surface over commissioned MW, contracted occupancy and schedule months; add a probability distribution for approvals, supply constraints and commissioning outcomes, then reconcile developer and tenant economics to avoid double counting.

09 — Vendor-side commercial opportunity map

A very large construction budget is a procurement context, not a sales pipeline. The qualifying chain is: defined project package → package owner and contracting party → confirmed technical requirement → tender/procurement method → procurement timing → eligibility → deliverable supplier capability → competition and price → financed execution → awarded order. Only then can a potential opportunity enter an expected-contribution model.

Relevant categories may include transformers and high-voltage equipment, protection relays, switchgear, optical fiber and physically diverse routes, cooling systems, industrial controls, power quality monitoring and commissioning services. Presence on this list does not establish a planned tender for any particular supplier, an available contract amount or an introduction to CloudHQ.

10 — Counterparty, operating and regulatory diligence

• Capacity: establish which MW are IT, facility input, import entitlement, contracted supply, tested demand and reserved electrical headroom.
• Commissioning: dated acceptance requirements; utility and customer test protocols; redundancy class; energized assets; fault-ride-through and protection coordination.
• Customers: executed lease or service agreement, initial billing trigger, ramp schedule, take-or-pay provisions, credits, termination rights and counterparty credit strength.
• Grid: interconnection agreement, grid-impact studies, queue status, firm/nonfirm conditions, right-of-way and curtailment/quality risks.
• Costs: vendor quotations, imported component content, FX, logistics, schedule contingencies, insurance, availability penalties and alternative sourcing.
• Environment: approved site permits and water/cooling specifications; distinguish advertised waterless cooling from independently measured lifetime water consumption.
• Finance: funding commitments and disbursements, lender conditions precedent, interest during construction, tax treatment and working-capital needs.

11 — Competing hypotheses and attempted falsification

Alternative A: there is no economically material delay because tenants pay only on a fixed later date independent of commissioning. Then early energization creates minimal cash-flow value. Alternative B: available power is not the binding constraint; tenant contracting or telecom backhaul is. Then acceleration spend on substations may have zero incremental effect. Alternative C: apparently inconsistent MW figures are simply phase distinctions. Then the conclusion is reconciliation, not misstatement. Alternative D: market demand supports the campus but phased rollout yields superior return to early scale. A real investment recommendation must test all four against documents before deciding.

12 — Quantifying LARGITAS's potential decision contribution

Deliver a dated capacity ledger, project-critical-path diagram, tender-package registry, supplier qualification matrix, contract-specific financial workbook, 5–95% sensitivity intervals and an executive go/no-go decision memo. The value claim should be expressed only as the result of a decision comparison: NPV of selected alternative minus NPV of defensible baseline after advisory fees and implementation costs. For example, a hypothetical US$0.994m expected value is not a LARGITAS-earned saving, nor a guaranteed client benefit; it is the output of an assumed decision problem.

13 — Failure conditions and next-verification sequence

Reject the acceleration proposal if the targeted bottleneck is not critical, if accepted capacity cannot be billed on the accelerated date, if contractor incentives are misaligned, if probabilities are unsupported or if all-in acceleration costs exceed probability-weighted NPV. Freeze inconsistent capacity figures as unresolved, commission a phase-by-phase reconciliation, obtain the commercial start-condition, then replace assumptions with executed contract and engineering schedule data. No vendor engagement or client relationship is implied by this independent company analysis.

14 — Publication discipline and reference register

[S1] CloudHQ QRO presentation, September 2025, especially page 4: 600 MW computing load, first 200 MW power, up-to-900 MW campus substation, 2 GW switching station and planned US$4.8bn construction (linked in primary sources below).

[S2] Mexican Presidency announcement, September 25, 2025: announced scope, location, investment and jobs. [S3] CloudHQ QRO webpage, accessed October 23, 2026: 360 MW display, six 48 MW descriptions, RFS within 2027 and 400 kV interface. [S4] El Economista, September 10, 2026: attributed executive remarks on land, energy and substation progress. [S5] El Economista and Data Center Dynamics, October 2026: reported updated proposed scale. [S6] Proyectos México electricity profile: published 2025 national system metrics. [S7] SENER electricity planning, via official publication and contemporary reporting: Querétaro constraints and proposed grid reinforcement. Figures are dated; announcements are not actuals.

15 — Additional diligence annex: critical-path dependencies

Before paying an acceleration premium, trace four interlocked critical paths independently: property/permits, grid interconnection, mechanical and electrical completion, and commercial tenant acceptance. A shorter equipment delivery schedule has zero financing value if the interconnection agreement or tenant start condition still governs first billing. In a staged campus, record the first building separately from the total campus to prevent capacity and CAPEX double counting.

Electrical definitions and engineering conversions

Apparent power S (MVA) and active power P (MW) are linked by P = S × power factor, within equipment and thermal limits. Facility input energy at one measured time can be approximated by IT consumption × verified instantaneous PUE; annual PUE or commissioning assumptions require their own metering basis. For example, a 100 MVA transformer at 0.95 power factor is nominally associated with 95 MW active power before reserves, derating, redundancies and system constraints; this is a dimensional illustration, NOT the rating of the QRO project. No public MW figure should be promoted to a definitive contracted supply entitlement without utility records.

Scope-change risk and October 2026 project updates

The proposed move from US$4.8bn toward nearly US$6bn in October 2026 is an increase in the reported project funding envelope of approximately US$1.2bn, or 25%, if comparing the two rounded announcements. It is NOT a statement of additional capital already spent, a verified cost overrun, a full financing commitment or additional signed customer demand. The later 'nearly 1,000 MW' refers to energy-infrastructure plans, not proof of 1,000 MW of billable IT. A serious analyst requests the dated change-control register, added scope and revised contract schedule. [S5]

16 — Risk-adjusted acceleration: full scenario grid

The intervention decision is not based on upside alone. P(success) must represent a probability of additional earlier operation caused by the spending, not the overall probability that the campus will someday complete. If commissioning date probabilities are correlated with tenant occupancy or grid congestion, the 3×3 grid alone overstates certainty and a scenario tree or simulation with correlated states is necessary. The displayed values are approximate rounded figures.

17 — Commercial sourcing and verification sequence

The logical sequence for a paid mandate is: freeze factual record at an as-of date; identify contract/electricity boundaries; validate engineering path and customer commencement terms; quantify baseline cash flows; model ranked feasible interventions; stress timing, power, tenant and financing risks; solicit necessary formal quotations with authorization; then report an evidence-weighted go/no-go recommendation. This firm must not imply that it possesses confidential client documents or has secured buyer introductions.

18 — Three-case assumption protocol

The three sets below are boundary tests, not a fitted probability distribution. Use the central reference as the working case and inspect downside/upside; never average scenario outputs without defensible weights.

Assumption disclosure: the 48 MW public building reference is factual; occupancy, contribution, marginal costs and success probabilities are scenario inputs, not CloudHQ contracts. Values follow the same 120-month annuity / three-month shift method as the central study. These success probabilities are illustrative, so EV is conditional rather than forecast.

Decision rule: report the CENTRAL scenario as the provisional working case and the full range separately. Do not take the mean of three hand-picked cases; use a probability-weighted expectation only when probabilities have an auditable empirical basis.

19 — Documented industry precedent: measured cooling-energy reduction, not a LARGITAS result

The U.S. EPA ENERGY STAR program records a completed 2013 air-side economizer retrofit at Marvell Semiconductor’s 720 kW data center in Santa Clara, California. The publisher lists mean monthly electricity use of 1,361,450 kWh in the preceding 12 months and 1,091,280 kWh in the following 12 months; difference 270,170 kWh/month. Annualized arithmetic at those reported monthly averages is 3,242,040 kWh/year, equivalent to 19.84% of the earlier monthly mean. ENERGY STAR reports US$27,000/month, or US$324,000/year, of savings, US$662,000 installed cost and a US$171,000 utility rebate. Recomputed simple payback: 662,000/324,000 = 2.04 years gross, or (662,000−171,000)/324,000 = 1.52 years after the rebate. [S9]

Attribution boundary: This is an externally documented historical intervention by another organization, not a CloudHQ campus result and not a LARGITAS client outcome. An observed before/after comparison is useful but not the same as a controlled experiment: weather, utilization, load growth and other operations may differ between the twelve-month periods. These numbers illustrate that facility engineering decisions can have quantifiable monetary consequences; no saving should be carried over mechanically to Querétaro without climatic engineering, tariffs, capital pricing, load shape and verified contract terms.