McAllen Data Center, LLC has to prove whether marketed data-centre capacity can survive power and carrier constraints

Summary

• McAllen Data Center, LLC, trading as MDC Data Centers, has a verifiable McAllen footprint: MCA2 at 422 South 11th Street, the legacy MCA1 site at 200 South 10th Street, a 2022 Texas renovation filing, ARIN resources and a current PeeringDB facility and exchange presence.
• MCA2's public specifications state 500 kW of electrical capacity with an upgrade path to 1 MW, 20 minutes of AC battery backup, two hours of DC battery backup, N+1 CRAC cooling, an emergency power plant, carbon-dioxide fire suppression, and 10G and 100G exchange ports. They do not disclose dual utility feeds, generator fuel runtime, UPS topology, concurrent-maintainability, flood protection or actual customer failover results.
• The McAllen interconnection case is stronger than the power case. PeeringDB lists 29 networks at the combined McAllen facility record and 19 connections at MEX-IX McAllen, while Axtel and GTT have publicly described moves into MCA2. Those records show active network use, but exchange port capacity and entity count do not prove that fibre routes, meet-me rooms or customer circuits have independent failure domains.
• The appropriate conclusion is split: Medium evidence that MCA2 is an active interconnection site, but Weak public evidence for facility resilience and customer recovery. Buyers should treat marketed capacity as usable only after receiving site-specific electrical one-lines, generator endurance, fibre-route drawings, maintenance terms and witnessed failover results.

The important number is not one megawatt

The most eye-catching number on the current MCA2 facility page is electrical capacity of 500 kW “to upgrade 1 mW.” Read generously, that means an operating base with a path to one megawatt. Read rigorously, it means neither one megawatt of installed critical load nor one megawatt available to new customers has been demonstrated. The phrase combines a present figure and a future possibility without stating how much utility service, switchgear, UPS capacity, generator capacity, cooling rejection, rack distribution and saleable inventory are available today.

That distinction matters more at MCA2 than it would at an ordinary enterprise server room. MDC presents McAllen as the place where Mexican carriers, international networks, content providers, the MEX-IX exchange and cross-border fibre meet. A single cabinet can therefore carry traffic whose economic reach is much larger than its power draw. The facility does not need to be hyperscale to be important. It needs to be dependable at every layer that connects a router's power supply to the outside world.

The company gives enough detail to establish a serious operating surface. It states 480, 277, 208 and 120-volt service, 20 minutes of AC battery backup, two hours of DC backup, N+1 computer-room air-conditioning units, an emergency power plant, carbon-dioxide fire suppression and exchange ports at 10G and 100G. Its wider McAllen location page names two sites: MCA2 Core at 422 South 11th Street and MCA1 Legacy at 200 South 10th Street. It says the international crossing links McAllen and Reynosa through dark fibre and identifies MCA2 as the location of MEX-IX.

Those are meaningful claims. They also show exactly where the unanswered questions begin. “Emergency power plant” does not reveal generator count, net usable output after cooling and ancillary loads, fuel type, on-site fuel quantity, refuelling priority or tested runtime under design load. “N+1 CRAC” describes unit redundancy but not cooling-loop, control, power-source or heat-rejection redundancy. Battery duration does not reveal battery age, end-of-life design, test regime or whether every customer circuit is covered. An upgrade path does not say whether permits, utility work, transformers and mechanical plant are complete.

The public evidence therefore supports neither a dismissal nor a blank endorsement. MCA2 is not an imaginary project attached to a corporate name. It is a named facility with live network records and public customers. But the capacity that matters is the capacity that remains powered, cooled, reachable and serviceable through a fault. That amount cannot be calculated from the published figures.

Two buildings became one operating centre of gravity

MDC's own pages draw a sharp boundary between old and current McAllen infrastructure. The MCA1 page calls 200 South 10th Street “MCA1 Legacy.” It says the McAllen connectivity community began there, but that the ecosystem now lives at MCA2, a purpose-built, MDC-owned site at 422 South 11th Street. It adds that all interconnection and international fibre crossings now land at MCA2.

That statement clarifies ownership and operations better than a generic location list. It identifies MCA2 as the present core and says MDC owns it. It also creates concentration risk. If the exchange, crossings and carrier community have been consolidated from a legacy building into one current core, the relevant question is no longer whether the company has two street addresses. It is whether those addresses provide independent service or whether MCA1 is now mainly historical. A second building only improves resilience when equipment, power, cooling, routes, staff and customer configurations can actually carry service there.

The legal identity is reasonably coherent. ARIN's registration for AS27286 identifies the organisation as McAllen Data Center, LLC and describes the autonomous system as MCALLEN-DATA-CENTER. ARIN's organisation record connects the same company name to the McAllen address and public network contacts. PeeringDB's organisation page gives the long name McAllen Data Center, LLC, the operating name MDC Data Centers, and a portfolio of McAllen, El Paso, Laredo, Nogales and San Diego facilities. A commercial company profile from Dun & Bradstreet also places McAllen Data Center, LLC at 422 South 11th Street and associates it with the MDC name and website.

Those records make the operator boundary credible, but they do not prove that every MDC-branded location has the same ownership, topology or resilience. Nor does a multi-city portfolio automatically fail over a McAllen exchange entity. A router connected only at MCA2 cannot move itself to El Paso because the operator owns or markets another building there. The customer needs a second port, transport, routing design, configuration and tested operational procedure.

Recent carrier moves make the MCA2 concentration more concrete. In January 2026, Axtel said MCA2 would become its principal new McAllen point of presence, completing its exit from the Chase building. In July 2026, MDC announced that GTT had consolidated its McAllen operations at MCA2, also describing a migration from the legacy site. These are strong signals that the facility is active and commercially relevant. They are simultaneously evidence that more traffic and more operator equipment now share the MCA2 failure domain.

Consolidation has real benefits. Networks can shorten cross-connects, find more counterparties in one meet-me environment, simplify remote support and reduce the cost of maintaining multiple small sites. It can improve performance and make a market commercially viable. Yet density also raises the consequence of a common-mode failure. A utility event, cooling-control fault, fire-suppression discharge, access restriction, meet-me-room incident or mistaken maintenance action at the core can affect parties that thought they were diverse because their carrier names were different.

The proper interpretation is not that consolidation is inherently unsafe. It is that the value of MCA2 creates a higher burden of proof. The more the operator calls one building the “true core,” the more customers need evidence that the core can survive predictable failures and that a failure inside it does not defeat every route sold as diverse.

The construction record confirms a site, not its full capacity

The strongest official facility record is modest and useful. The Texas Department of Licensing and Regulation project filing TABS2022015724 names the project “McAllen Data Center,” the facility “MDC Data Center,” and the location 422 11th Street in McAllen. It records privately funded renovation work scheduled from May to July 2022, an estimated cost of $200,000, a scope of minor interior demolition and renovation to an existing data-centre building, and 4,432 square feet of affected area. The listed owner is MDC Data Center at the same address.

This record does three things well. It independently confirms physical work at MCA2's address, ties the work to the operator name and shows that the 2022 project was an alteration of an existing building rather than a greenfield campus. It does not certify the resulting electrical capacity, generator design, cooling topology or completed operating state. The page marks the project as registered and explains the accessibility inspection obligation; it is not a commissioning report for critical infrastructure.

A third-party datacenters.com listing for MCA2 describes a 16,000-square-foot building and 1 MW of power. Another third-party listing reports two 500 kW generators in an N+1 arrangement. These details are plausible, but they should not be fused into a single guaranteed specification. The Texas filing's 4,432 square feet may describe only the renovation area, while 16,000 square feet may describe the whole property. The operator's own current page says 500 kW with an upgrade path to 1 MW, while the marketplace page presents 1 MW as total power. Without a dated facility data sheet and electrical one-line, it is not clear whether the difference reflects design capacity, completed capacity, utility capacity or simple profile lag.

This is a common data-centre problem: numbers with different denominators appear equivalent. Building area is not white space. White space is not fitted cabinet space. Utility capacity is not critical IT capacity. Generator nameplate capacity is not generator-backed customer capacity. Design capacity is not installed capacity. Installed capacity is not uncommitted inventory. A buyer that asks only “How many megawatts?” can receive a technically true answer that is commercially misleading.

MCA2's stated 500 kW also places it far below the giant artificial-intelligence campuses driving current Texas power debates. ERCOT's large-load integration material is aimed at loads measured in tens of megawatts and, under the 2026 batch framework, projects of 75 MW and above. That statewide context is relevant to utility planning, but it should not be used to exaggerate MCA2's scale. A sub-megawatt interconnection site can still be locally constrained by transformer availability, feeder configuration, building switchgear, generator endurance and cooling.

The evidence needed to settle the capacity question is straightforward: a dated owner-issued facility specification; the current utility service rating; the critical-load rating after mechanical overhead; installed and commissioned UPS modules; generator net output and runtime; occupied and available cabinet power; and the work remaining before the advertised upgrade can be sold. Until those values align, one megawatt is a direction, not a demonstrated customer resource.

Power resilience is described at component level, not system level

McAllen sits within the AEP Texas service area. AEP's McAllen and Mission transition page says that in November 2019 it acquired the former Oncor distribution assets in and around those cities. AEP's current tariff identifies McAllen in its certified service area. That establishes the distribution utility context, but it does not identify MCA2's feeder, substation, utility service count or priority during restoration.

MDC does not publicly state that MCA2 has dual utility feeds. That omission matters because “redundant power” can refer to redundancy downstream of one utility service. A facility may have multiple UPS modules, multiple distribution panels and multiple rack feeds while still depending on one outside feeder or one service entrance. Such a design can be robust against equipment failure inside the building and still be exposed to a utility fault before the service entrance.

The generator layer is similarly incomplete. The MCA2 page names an emergency power plant but gives no runtime. A commercial listing says there are two 500 kW generators, yet even that figure does not tell a customer how the plant behaves. Are both generators required to carry the design load, leaving no spare unit? Does N+1 apply at current 500 kW but disappear after a 1 MW upgrade? Is cooling fully generator-backed? Are fuel pumps, controls, security, meet-me rooms and office loads on the same protected bus? How often are generators tested at load, and how is fuel replenished during a regional storm?

The battery figures need the same care. Twenty minutes of AC battery backup is enough to bridge a normal generator start, but it is not a promise of twenty minutes at every load, battery age and ambient condition. Two hours of DC battery backup may be valuable for carrier equipment designed around -48 VDC, but it does not necessarily support ordinary AC servers or cooling. If utility power and generators are unavailable, DC routers can remain electrically alive while the room warms and connected optical systems elsewhere fail. Battery duration is one part of a timed sequence, not an availability result.

Regional outage history shows why generator endurance cannot remain a brochure footnote. Following severe Rio Grande Valley storms in April 2023, AEP reported that 155,000 customers had lost service and 126,000 had been restored by Sunday evening; it still listed restoration targets by area. In March 2025, extreme rain in the border region caused deaths, rescues, school closures and thousands of power interruptions, according to Associated Press reporting. These events do not prove that MCA2 went offline. They prove that a serious local facility must be designed for outages and access conditions that can last much longer than a battery bridge.

Fuel is an operating dependency as much as an engineering specification. A generator that can run for a stated number of hours assumes a starting tank level. A refuelling plan assumes roads are open, vendors are available, contracts are honoured and the fill point is accessible. A customer assessing MCA2 should ask for runtime at the current critical load, minimum fuel policy, refuelling contracts, post-event priorities and the most recent full-load test. It should also ask whether generator maintenance ever removes the site's spare capacity and whether customers receive notice.

There is no public incident history showing a failed utility transfer, and it would be wrong to invent one. The downgrade comes from missing proof, not proof of failure. MDC has disclosed components consistent with a real data centre. It has not disclosed the system-level evidence needed to know whether those components preserve every critical service through a long outage.

Cooling defines the usable rack envelope

MCA2 lists N+1 CRAC cooling. That is better than listing no redundancy, but N+1 is meaningful only after the load and failure boundary are defined. One spare CRAC unit can cover the loss of one active unit. It cannot necessarily cover a common control failure, a lost electrical panel, a blocked condenser, a failed heat-rejection loop or a sustained heat wave that reduces unit performance. If every CRAC depends on the same upstream power or control path, unit count can overstate resilience.

Cooling also determines whether the 500 kW electrical number can become IT load. Some facility power is consumed by cooling, fans, pumps, lighting, security, conversion losses and support systems. The amount left for racks depends on environmental design and operating efficiency. A small interconnection facility may contain many network cabinets with moderate average draw rather than dense computing racks, but that does not remove the thermal constraint. Routers and optical transport systems can be heat-sensitive, and a crowded meet-me environment can create local hot spots.

McAllen's climate makes the question practical. High summer temperatures and humidity increase cooling demand and can narrow the margin available during equipment failure. Regional storms can combine a utility loss with difficult outdoor conditions and restricted staff access. A facility may transfer successfully to generators and still face a cooling incident if mechanical starting loads, controls or condenser systems do not behave as designed.

The published specification does not state rack density, temperature and humidity limits, cooling runtime on generator, containment arrangement, sensor coverage or the capacity available after the largest cooling component fails. It also does not state whether the claimed upgrade to 1 MW includes a corresponding mechanical upgrade. Electrical expansion without heat-rejection expansion is not customer-usable growth.

A serious buyer should ask for cooling capacity in normal and failure states, not just CRAC count. It should request the current critical cooling load, N+1 capacity at design outdoor conditions, generator-backed mechanical scope, alarm escalation, maintenance history and evidence from a controlled unit-failure test. It should also ask how long customer equipment can remain within environmental limits if cooling is lost while power remains available.

Again, the answer may be satisfactory. MDC markets 24/7 engineering support, scheduled activities, monitoring, equipment replacement and on-demand remote or smart hands. Those services suggest staff and operational procedures exist. The page does not give MCA2 staffing levels, guaranteed response times for each service tier or cooling-recovery results. Support availability is therefore a positive operating signal, not a substitute for mechanical proof.

Fire and flood can defeat otherwise redundant systems

The MCA2 page states carbon-dioxide fire suppression. Carbon dioxide can suppress fire without the water damage associated with sprinklers, but the public line is too brief to establish the protected zones, detection method, discharge logic, life-safety controls or interaction with code-required systems. A suppression event can itself cause service interruption through evacuation, equipment shutdown, access restrictions or accidental discharge. Customers need to know whether separate rooms and meet-me areas are isolated into fire zones and whether one event can affect the whole operating floor.

The 2022 Texas renovation filing confirms building work but does not expose the fire-system commissioning evidence. Nor does the public facility page state whether the electrical rooms, generator area, batteries, carrier meet-me area and customer space have separate detection and suppression. A one-line label should not be treated as a complete fire strategy.

Flood exposure deserves equal attention. McAllen and the wider Rio Grande Valley have repeatedly experienced heavy rain and flash flooding. The City of McAllen's strategic plan describes ongoing storm-system inspection, flushing and storm-sewer repair intended to reduce flood likelihood, including work during hurricane season. The city's own Hurricane Hanna damage-assessment map service records flood-response layers created after the 2020 storm. These sources establish a regional hazard; they do not establish the flood elevation of MCA2.

The March 2025 border rain event reinforces the point. Flooding can interrupt more than the building itself. It can block technician access, delay fuel deliveries, expose underground vaults, damage utility equipment and prevent carriers from reaching repair points. A building can sit outside a mapped high-risk zone and still lose service through a route, substation or street-level dependency elsewhere.

MDC does not publicly state MCA2's finished-floor elevation, flood-zone classification, generator and fuel-system elevation, watertight entry design, sump capacity or below-grade fibre-vault protections. That absence does not prove unusual exposure at 422 South 11th Street. It means the site-level hazard remains unresolved in public.

The evidence that would settle it is specific: current flood-map determination; finished-floor and critical-equipment elevations; drainage and water-entry controls; location of utility and fibre vaults; fuel-system protection; emergency access routes; and the operator's experience during Hurricane Hanna and the March 2025 rains. Customers should also ask whether flood and hurricane exercises include carrier contacts and fuel suppliers, not only building staff.

Network density is real, but diversity is not a logo count

The public network evidence is the strongest part of the MDC case. PeeringDB's MDC McAllen facility record combines the MCA1 and MCA2 addresses, labels the facility with both site names, and lists 29 networks, two carriers and one exchange. The networks include Akamai, Alestra, Arelion, Axtel, Bestel, C3ntro Telecom, CDN77, Cogent, Gcore, GTT, Hurricane Electric, IENTC, Megacable, Tata Communications, Sparkle, Telxius, Totalplay and Zayo. This is not evidence of an empty room.

PeeringDB's MEX-IX McAllen page showed 19 connections, 19 peers and 1.4 Tbps of aggregate listed port capacity during this review. It publishes the IPv4 exchange prefix 206.71.142.0/24 and IPv6 prefix 2620:3b:6000::/64, and identifies MDC McAllen as its local facility. Entity entries include major Mexican access networks and international content or backbone providers. The Internet Society Pulse tracker independently reflects the exchange and its entities using PeeringDB-derived information.

The distinction between listed port capacity and traffic is crucial. Adding the nominal speeds of connected ports produces an aggregate capacity number. It does not show average traffic, peak traffic, committed usage, oversubscription, packet loss or the traffic that would remain after a port, switch or site failure. MDC's own November 2025 announcement said MEX-IX exceeded 500 Gbps across McAllen and El Paso. That is a meaningful scale claim, but it combines two cities and comes from the operator. It should not be read as a third-party measurement of current McAllen traffic.

MDC's MEX-IX service page says the platform offers 10G and 100G ports, bilateral peering, route-server access, a 24/7 network operations centre and a carrier-grade service commitment. PeeringDB identifies AS27420 as the McAllen route-server network and AS27286 as the route collector. Those are operationally coherent records. The AS27286 PeeringDB entry lists no originated customer prefixes because a collector observes routing rather than serving as evidence of a transit customer base. An AS number tied to the company must not be mistaken for proof that the facility itself has multiple upstream routes.

Carrier neutrality also needs precise interpretation. A neutral facility gives customers a choice of networks and cross-connects. It does not guarantee that any two chosen carriers enter through physically independent conduits, use different long-haul routes, terminate on separate meet-me-room equipment or avoid a common border crossing. Two logos can share duct, bridge, street, regeneration hut, power supply or upstream backbone. Commercial diversity is not automatically physical diversity.

MDC's interconnection page markets express cross-connects, patching, troubleshooting, MEX-IX access and international crossings. These are relevant services, but the public pages do not disclose cross-connect installation intervals, meet-me-room count, diverse entry points at MCA2, spare switch capacity, exchange fabric topology or maintenance performance. PeeringDB's combined MCA1/MCA2 facility record also makes it difficult to know which listed networks are physically present at which building without asking the operator or the network.

The network grade is therefore Medium rather than Strong. The exchange, prefixes, route servers, facility records and entity announcements are visible and current enough to support active use. What is missing is independent route geometry and failure evidence. Customers should request a site-specific carrier list, exact building and room for each handoff, entrance path, conduit ownership, shared-risk groups, exchange switch redundancy, port allocation and the result of the latest fabric failover test.

Cross-border fibre is valuable and permit-dependent

MDC's strategic advantage is not simply a room in McAllen. It is a room close to Mexico with fibre routes intended to cross the border. The McAllen page says its international crossing connects McAllen to Reynosa through dark fibre. The broader international fibre crossing page markets nine crossings across Texas, Arizona and California, diverse routes, leases and 15- or 20-year indefeasible rights of use. It says customers can use their own equipment rather than buy a third-party lit service.

That operating model can give carriers more control. Dark fibre lets a network select optics, capacity and upgrade timing. A long-term right can support investment in route-specific equipment. Multiple crossings can reduce dependence on one border city. But none of those portfolio-level advantages proves that two circuits ordered at MCA2 avoid the same local path between the rack, building exit and border.

The physical route includes many possible control points: the customer's router, cross-connect, meet-me frame, building entrance, street duct, private easement, public right of way, bridge or river crossing, customs or security access, Mexican landing point, power for optical equipment and the onward route in Reynosa. A “diverse” international crossing may still share the MCA2 meet-me room or a short local conduit before separating. Conversely, two services sold by different carriers may become genuinely independent after a shared building segment. The route drawing, not the product name, decides.

Public-government records show why permits and leases belong in the availability analysis. In El Paso, a city council record documents a lease to McAllen Data Center, LLC for a duct in the city's underground conduit system serving an international fibre system near the Zaragoza bridge. A later council resolution adjusted the depicted pathway. That record concerns El Paso rather than McAllen, so it cannot establish MCA2's route. It does demonstrate the type of municipal right, route definition, annual obligation and physical public infrastructure that cross-border fibre can depend on.

The McAllen-Reynosa crossing needs its own equivalent evidence: landing points, route owner, permits, maintenance rights, crossing structure, strand allocation, repair authority and physically separate alternatives. The public marketing page does not provide those details. Nor does it state average repair time for a river or bridge crossing, availability of spare fibre, optical regeneration points or emergency access on both sides of the border.

An international route also has two-country operating exposure. A cut on the US side and a cut on the Mexican side may involve different contractors, authorities and access conditions. A dark-fibre customer controls the optical layer but may still depend on MDC or another owner for physical repair. The contract should identify who detects a loss, who opens the repair, who has test equipment at each end, who crosses jurisdictional boundaries and whether another route is already configured.

The right conclusion is that MDC has a credible cross-border infrastructure proposition, not that every marketed crossing is proven resilient. The facility's importance follows from the route opportunity. Its evidence burden follows from the same fact.

Consolidation changes who is affected by failure

A failure at MCA2 would not automatically disconnect all listed networks or all Mexican users. Large carriers generally operate wider backbones, and many exchange entities have other interconnection points. The impact depends on what each entity places in McAllen, which routes it announces there, whether traffic can reconverge elsewhere and whether the failure affects colocation, exchange, private cross-connects, fibre crossings or all of them.

Still, public announcements identify several potentially affected groups. Axtel says MCA2 became its principal new McAllen point of presence. GTT says it consolidated McAllen operations there. PeeringDB lists Mexican operators, content providers and global carriers at the facility and exchange. MDC says the site hosts bilingual engineers and its border interconnection platform. These facts support a reasonable impact mechanism: a common MCA2 event could remove local peering sessions, private handoffs or border transport until networks reroute or technicians restore service.

The first affected group would be network operators with equipment or ports in the building. They could lose reachability to local peers, face increased transit cost, send traffic over longer routes or lose a border circuit. The second group would be content and cloud networks using McAllen to deliver traffic into Mexico. Their users might experience higher latency or congestion if traffic shifts to another city. The third group would be Mexican access networks collecting content or exchanging routes at MEX-IX. The fourth would be enterprise or wholesale customers whose private service crosses at McAllen.

The severity would vary. Losing an exchange port may be a performance event if transit remains available. Losing a unique cross-border circuit may be a full service outage. Losing power to customer routers while the exchange fabric stays healthy may affect only some entities. Losing the meet-me environment or a shared fibre entrance could affect ostensibly separate services. A credible incident plan has to distinguish these cases quickly.

MDC's 24/7 support and customer portal may help coordinate response. But public material does not show incident notices, restoration statistics, a status history or examples of customer failover. The absence of a public incident archive is not evidence that incidents were mishandled. It prevents outsiders from measuring communication speed, root-cause quality and recovery performance.

Customers should ask for anonymised evidence: recent availability reports, maintenance notices, transfer-test results, exchange-fabric failover records, mean repair times, post-incident summaries and references from customers with comparable designs. They should also define the business consequence of each failure in advance. A network buying a low-cost peering port has a different recovery objective from a carrier using MCA2 as the only handoff for a regulated or safety-relevant service.

Maintenance is a hidden capacity constraint

Physical redundancy is only useful when equipment can be removed from service without exhausting the spare margin. This is the practical meaning behind concurrent maintainability. A facility can have N+1 cooling and two generators on paper, yet lose resilience whenever one unit is under repair, one UPS module is bypassed or a switch is being upgraded. If customers continue to be sold up to the normal limit during those windows, the nominal spare component may no longer be spare.

MDC publishes service options for scheduled work, monitoring, inventory, installation, replacement and on-demand support. That indicates an active maintenance operation. It does not state MCA2's standard maintenance windows, customer notice periods, blackout dates, escalation rights or whether the electrical and cooling systems can be maintained without exposing single points of failure.

The move from MCA1 to MCA2 is itself a useful operating test, but public announcements describe the commercial result rather than the change procedure. A migration of Axtel or GTT equipment can involve parallel circuits, temporary cross-connects, route changes, power work and coordinated cutovers. Evidence that those migrations completed successfully would support staff capability. It would not prove generator, utility or cooling failover, because a planned network migration is a different test.

Human error must be included in the failure model. Cross-connects can be patched incorrectly. Breakers can be opened under the wrong change ticket. Exchange configurations can withdraw routes. Fire systems can be placed in the wrong mode. A redundant device can fail to take load because it was not maintained or because two supposedly separate systems share a control dependency. Mature operators reduce these risks through change review, labelling, access controls, rollback plans, dual authorisation and rehearsed emergency procedures.

None of that should be inferred from a promise of remote hands. A buyer should ask how changes are approved, whether customer work requires a second-person check, how emergency work differs from planned work, what notice applies, how maintenance risk is communicated and whether post-change validation includes power and route checks. It should also ask whether support subscriptions receive different response priority during a regional event.

Maintenance affects saleable capacity too. If the 500 kW figure is close to the installed limit, taking a power module or generator out of service may require load restrictions. If the 1 MW upgrade is under way, construction can introduce dust, accidental damage, temporary cooling changes and access conflicts. Current customers need to know whether expansion work changes their risk and what barriers separate live space from construction.

The public record does not show a current expansion project beyond the operator's upgrade phrase. That uncertainty should remain explicit. The issue is not that construction is known to be late or unsafe. It is that the upgrade status, critical path and effect on usable inventory are not published.

Installed capacity is not customer-available capacity

The capacity test can be reduced to a sequence. First, utility and generator power must reach the building. Second, switchgear and UPS systems must convert and distribute it. Third, cooling must remove the heat. Fourth, racks, power strips and cross-connects must be installed. Fifth, fibre and exchange ports must have spare capacity. Sixth, staff must be able to provision and repair the service. Seventh, the customer must have a configuration that uses the available redundancy.

Any one of those steps can set the limit. A facility with 500 kW of utility service but 350 kW of cooling-supported critical load cannot sell 500 kW of IT load. A room with spare power but no rack space cannot sell cabinets. A meet-me room with spare switch ports but no diverse fibre entrance cannot meet a customer's route requirement. A building with two carriers does not give a customer failover unless the customer orders, configures and tests both.

MCA2's public numbers do not reveal the limiting step. The operator may have ample uncommitted room, or the core may be commercially dense. The listed networks show ecosystem depth but not cabinet occupancy. MEX-IX port capacity shows possible interconnection throughput but not spare switch fabric. The stated path to 1 MW shows expansion intent but not energisation.

This is why buyers should request a capacity schedule with dates and definitions. It should separate utility service, generator-backed critical capacity, UPS-backed AC, DC plant, mechanical capacity, installed rack power, contracted load, measured peak load and immediately saleable load. It should state whether the “upgrade” is designed, permitted, ordered, installed, commissioned or available. Those words represent different investment stages.

The same schedule should identify network inventory: available 10G and 100G ports, cross-connect lead times, fibre-pair availability, handoff locations and any constrained carrier. If a customer requires two physically diverse paths, the operator should confirm the separate entrances and shared-risk groups in writing. If a customer requires a second MDC city, the quote should include transport and active failover rather than treating a portfolio map as redundancy.

MDC's commercial strength is that these elements can be bought in one place: colocation, support, exchange access and border crossing. The risk is that a bundled offer can hide the weakest layer. A cabinet may be ready while the diverse route is not. A fibre pair may be available while additional critical power is not. Honest capacity is the minimum of all required layers.

What customers should require before treating MCA2 as resilient

The first requirement is a site-specific responsibility map. It should identify McAllen Data Center, LLC as landlord, facility operator, exchange operator and fibre owner where applicable, and identify any other entity responsible for utility delivery, cross-border repair, Mexican landing infrastructure, carrier transport or customer equipment. A sales brand should not blur who has authority during an outage.

The second is an electrical evidence package. At minimum, it should show the utility service count and route, one-line diagram, UPS topology, battery design duration, generator configuration, generator-backed cooling, fuel capacity, runtime at current peak load, maintenance state and latest load-bank and transfer tests. It should explain how redundancy changes at 500 kW and at the proposed 1 MW state.

The third is a cooling package. Customers need normal and failure-state capacity, design outdoor conditions, CRAC power sources, control dependencies, generator-backed operation, rack-density limits, sensor coverage and recent unit-failure results. The promised power per cabinet should be supportable during the loss of the largest cooling component, not only on a mild day with every unit available.

The fourth is route evidence. A customer should receive carrier names at MCA2 rather than at the combined McAllen record, building entrances, meet-me-room paths, local conduit routes, border-crossing routes and known shared-risk groups. It should verify that a second carrier does not share the segment the customer is trying to protect against. For exchange resilience, it should ask which switches and power domains serve each port and whether a second port lands on a separate failure domain.

The fifth is recovery evidence. The contract should state maintenance notice, support response, escalation, remote-hands authorisation, replacement-part handling, generator refuelling, incident communication and service credits. More important, the design should be tested. Customers should witness utility transfer where practical, verify dual power feeds to equipment, pull a peering or transit path, test route reconvergence and confirm application recovery from the location meant to carry failover.

The sixth is flood and fire evidence. Finished-floor elevation, critical-equipment elevation, water-entry controls, flood determination, fire zones, detection, suppression, evacuation and restoration procedures all matter. The customer should understand whether a suppression event or water entry in one room can remove both of its supposedly redundant services.

The seventh is an exit design. Network operators should know whether they can keep addresses, how quickly cross-connects can be moved, how equipment is removed during an emergency, and whether remote hands can assist when travel is impossible. An interconnection site's value comes from density, but density can create lock-in when every relationship must be moved at once.

These requests are not an accusation. They are the normal evidence needed to convert component claims into an availability decision. A capable operator should be able to answer most of them under appropriate confidentiality.

The evidence grade must remain split

McAllen Data Center, LLC is easier to verify as an operating interconnection company than the initial thin corporate footprint might suggest. Official renovation records confirm work at MCA2. ARIN ties the legal name to network resources. PeeringDB shows a current facility, exchange, route collector, route server and a substantial list of networks. Axtel and GTT publicly identify MCA2 as their McAllen consolidation point. The company maintains detailed service pages and 24/7 published contact points. These facts justify a Medium network evidence grade.

The resilience grade is Weak. The public record does not provide dual utility evidence, a current electrical one-line, generator runtime, fuel policy, UPS topology, critical-load availability, cooling capacity under failure, current flood protection, separate fire zones, building-specific carrier routes, shared-risk groups, maintenance performance or customer failover results. The records show components and commercial use, not end-to-end survival.

That split is the most useful conclusion for customers and investors. It avoids the mistake of treating a small facility as unreal because it lacks hyperscale disclosure. It also avoids the more dangerous mistake of treating a live exchange and a long carrier list as proof that the building cannot fail. MCA2 can be both genuinely important and insufficiently transparent.

The company's strategic position is clear. McAllen is a practical meeting point for US and Mexican networks, and MDC has spent years concentrating operators, exchange services and crossings there. Its own pages now describe MCA2 as the core rather than merely one location among several. That concentration can produce lower latency, simpler interconnection and stronger network economics.

But concentration changes the standard. The facility must prove that its power, cooling, fire protection, access, meet-me environment and border routes do not collapse into one shared dependency. It must show that the stated upgrade can be energised and cooled, that emergency generation lasts through regional disruption, that diverse carriers are physically diverse, and that customer configurations actually fail over.

Until that evidence is available, the responsible reading of MDC's capacity is narrow. Five hundred kilowatts is a stated facility input, not a guaranteed customer outcome. One megawatt is an upgrade ambition, not verified present inventory. Twenty minutes of AC battery is a bridge, not a recovery plan. N+1 CRAC is a component arrangement, not proof of continuous cooling. Nineteen exchange connections and 29 facility networks show relevance, not immunity.

McAllen Data Center, LLC has proved that networks meet at MCA2. It has not yet proved in public how those networks keep meeting when the utility, cooling plant, carrier entrance or building itself is under stress. For infrastructure whose value comes from being a border crossroads, that is the proof that matters most.