A 600 MW AI site cannot establish power resilience simply by contracting for three separate incoming feeders. Each connection may have its own cable, breaker, and metering point while remaining exposed to the same transmission corridor, source substation, or control failure. That arrangement creates the appearance of redundancy without necessarily providing three independent ways to deliver power. The engineering question is not how many connections cross the site boundary, but how many can continue operating when a shared component fails. Designers must examine the entire electrical path, from the upstream grid source to the campus distribution system, before assigning any resilience value to a connection. For a site that depends on continuous high-density computing, independence must describe the physical infrastructure, electrical architecture, and control arrangements that determine whether the remaining supply can withstand a credible failure.

The requirement becomes more demanding when each path carries a planned 600 MW capacity, because the consequences of a shared failure extend beyond an ordinary feeder outage. A connection’s nominal rating does not establish the power that the complete route can deliver under actual operating conditions, and three nameplate ratings do not prove that the site can survive the loss of two routes. Engineers must establish the capacity available at each operating stage, account for restrictions elsewhere in the network, and verify that switching arrangements preserve the intended separation. They must also define the precise failure conditions against which independence will be tested, including faults, maintenance outages, protection operations, and loss of control power. Those tests need to examine both individual component failures and common-cause events that can affect multiple paths simultaneously.

Three Cables in One Trench Is Still One Failure Mode

Three incoming feeders can follow separate electrical circuits while occupying the same physical corridor, leaving them vulnerable to a single external event. A cable fault, excavation accident, fire, or flooding incident can affect multiple circuits when their routes converge within the same trench or duct bank. Separate cable insulation and individual circuit breakers cannot prevent a shared physical event from damaging all affected connections. Engineers must therefore evaluate duct bank separation, underground crossings, access chambers, and joint bays as part of the overall supply architecture. A joint bay that serves multiple routes can create a concentrated exposure during installation, maintenance, or an equipment failure. The relevant design evidence must show how each route avoids shared hazards, rather than merely documenting that three electrically separate feeders exist on the drawings.

Overhead infrastructure creates a comparable problem when multiple circuits depend on the same transmission towers or a common right-of-way exposed to the same hazards. A tower collapse can damage multiple circuits simultaneously, while severe weather can affect several nearby routes even when they use separate support structures. Underground and overhead sections also require a combined assessment because a route that remains separate for most of its length may converge at a critical crossing or transition point. Designers should map each feeder from its source to the site, identify every shared corridor and structure, and evaluate the consequences of losing those shared assets. Physical separation distances must reflect the local terrain, credible hazards, construction methods, and available alternative routes rather than an arbitrary distance chosen for presentation purposes. Route independence becomes credible only when the design demonstrates that one localized incident cannot remove multiple supply paths through the same physical exposure.

Your Independence Ends Where Your Fence Ends

A campus boundary provides a convenient point for measuring incoming capacity, but it does not define the limits of electrical independence. Three feeders entering through separate gates may originate from the same source substation, share an upstream transmission line, or depend on a common bus arrangement that exposes them to one fault. Two campus substations also do not establish independent supply if both receive power through the same upstream corridor or depend on the same vulnerable grid connection. Engineers need to trace each path through the transmission network to the source infrastructure and identify where the paths converge. That review must include source transformers, bus sections, circuit breakers, line terminals, and the upstream facilities required to deliver the contracted capacity. Without this wider view, a campus can count three incoming connections while overlooking a single upstream asset capable of interrupting all three.

Genuine source separation requires more than assigning different names to substations or connecting feeders to different terminals. The engineering team must establish which transmission circuits supply each substation, whether those circuits share structures or corridors, and how the upstream bus configuration responds to faults and maintenance. Separate source substations can still share a common generation constraint or transmission bottleneck, so physical separation alone does not guarantee that all three routes can deliver their intended output. A credible assessment should identify common-mode exposures, document the expected response to relevant contingencies, and confirm that surviving paths retain sufficient capacity under the resulting network conditions. Utility planning information and site-specific power-flow studies can help establish what remains available after a line, transformer, or bus section becomes unavailable.

One Brain Controlling Three Feeds Is Not Independence

Separate conductors and substations cannot deliver true independence when a shared control or protection failure can disconnect all three supply paths. Protection relays detect abnormal electrical conditions and initiate the actions needed to isolate faults, but their settings, communication links, and trip logic can create dependencies across otherwise separate circuits. A common protection panel, shared auxiliary power supply, or incorrectly coordinated trip scheme may extend the consequences of one event beyond the equipment that originally failed. The risk also increases when multiple feeders depend on the same switchgear assembly, shared bus section, or common breaker-failure protection arrangement. Engineers must examine the actual wiring diagrams, trip matrices, relay settings, and failure responses instead of assuming that separate circuit labels imply separate protective behavior. Each path needs a clearly defined protection boundary that limits a fault’s impact without compromising the protection required elsewhere in the electrical system.

Digital controls introduce another layer of dependency because several independent electrical routes may rely on the same supervisory platform, communication network, or control-power source. Shared supervisory control and data acquisition systems can create a common operational vulnerability, although a supervisory outage does not automatically mean that all feeders will trip. The critical question is whether the loss or malfunction of a shared control component can initiate unwanted switching, prevent required isolation, or remove the information operators need to manage the surviving supply. Communications routes deserve the same scrutiny when fiber cables, network switches, or overhead communication conductors follow the same physical corridor as the power infrastructure. Independent protection functions, appropriately separated auxiliary supplies, tested fail-safe behavior, and resilient communications can reduce these dependencies when the design supports them.

The Moment Two Fail and One Has to Carry 600 MW

The proposed transition from three 200 MW paths to two 300 MW paths and finally one 600 MW path exposes the difference between installed capacity and deliverable capacity. In a three-path arrangement, each route may normally carry 200 MW, but losing one route leaves 400 MW of nominal capacity across the other two. Losing a second route leaves only one 200 MW path, unless the surviving connection can support a higher transfer level than its normal operating allocation. A two-path arrangement carrying 300 MW per route faces a comparable limitation because one surviving 300 MW route cannot automatically sustain a 600 MW site load. The arithmetic describes the intended operating model, not proof that the equipment, upstream network, and campus distribution system can support each transition.

A final-source requirement of 600 MW therefore needs a specific engineering definition and a demonstrated operating response. The surviving route must have adequate capacity throughout its complete electrical path, including source transformers, transmission facilities, incoming switchgear, and downstream distribution equipment. Protection coordination must isolate the failed paths without tripping the healthy route, while switching procedures must allow operators to establish the required configuration within an acceptable time. The campus must also establish whether its critical load can tolerate the transition, because a source capable of delivering 600 MW after switching does not guarantee uninterrupted power during the switching interval. Load shedding, staged recovery, and temporary operating restrictions may form part of the design, but the owner must define their limits rather than treating them as invisible exceptions.

Genuine Independence Is Proven, Not Counted

A credible independence claim requires evidence that connects the site’s electrical design to the physical and operational realities of the supply network. Designers should produce route drawings, substation one-lines, protection schematics, control-power diagrams, and documented contingency studies that expose shared infrastructure across all three paths. Those records should identify common structures, upstream bottlenecks, shared switching equipment, communications dependencies, and any operating assumptions that limit the surviving supply. Commissioning should then test the intended protection boundaries, switching sequences, control failures, and recovery procedures against the defined design scenarios. Where a failure cannot be reproduced safely under live operating conditions, engineers should use appropriate simulations, secondary injection, functional testing, and other validated methods to establish the expected response.

Utilities need a clear statement of the capacity and network conditions they can support, while designers must translate those commitments into a verifiable site architecture. Insurers and customers need evidence that the proposed resilience matches the consequences of losing power, rather than relying on a feeder count or a simplified capacity diagram. However, absolute independence from every possible shared influence across a connected grid is rarely a practical engineering guarantee, so the requirement must specify credible failure scenarios, acceptable residual risks, and measurable performance criteria. Each party should agree on those criteria before construction and retain the test results, assumptions, and operating limitations as part of the site’s technical record. Three 600 MW paths become genuinely independent when engineers can demonstrate how each survives the defined failures without depending on infrastructure that those same failures can remove.