The Power Grid Is Running Out of Room for Its Backup Plans

PJM's electricity grid serves approximately 67 million people across the Mid-Atlantic and Midwest. Rising demand, generation shortages and transmission constraints are placing increasing pressure on the resources used to maintain reliability. (Patrick Lewis/Loudoun Now)

On October 9, federal regulators rejected a proposal that would have allowed the nation's largest electricity grid operator to impose special curtailment obligations on certain large electricity customers, including data centers. Three days later, the Department of Energy authorized that same operator to use emergency generating resources and, under specified conditions, direct large electricity users to operate their own backup generators. By October 13, PJM Interconnection had issued a Maximum Generation Emergency/Load Management Alert, recalled active maintenance outages and identified generating resources it needed available to maintain reliability.

The decisions weren't responses to a single event. The Federal Energy Regulatory Commission was addressing how PJM could accommodate rapidly growing electricity demand without compromising future reliability. The Department of Energy was responding to an immediate operating problem involving elevated demand, scheduled generation outages and transmission constraints. But both decisions concern the same electricity system, and both involve arrangements intended to preserve service when ordinary available resources may be insufficient. The distinction between the two problems matters, particularly because a future capacity shortage isn't the same as an immediate shortage of electricity. What makes their convergence significant is that PJM is having to address both while the resources available to compensate for either are becoming more consequential.

PJM coordinates wholesale electricity markets and regional transmission across all or parts of 13 states and the District of Columbia, serving approximately 67 million people. Its July capacity auction, intended to secure electricity resources for the 2028–2029 delivery year, obtained 138,318 megawatts of accredited generation and demand-response capacity. Including resources committed outside the centralized auction, PJM still fell 6,831 megawatts short of its reliability requirement. The auction price reached its federally approved ceiling of $325 per megawatt-day, marking the second consecutive auction that failed to secure the required capacity. The previous shortfall had been approximately 6,517 megawatts. Higher prices were encouraging additional supply, but not enough to keep pace with the resources PJM expected to need.

That shortfall doesn't mean PJM will be unable to provide electricity in 2028. Capacity requirements include reserve margins designed to protect against unexpected failures and periods of unusually high demand, and the system has additional opportunities to secure resources before the delivery year begins. The significance is that the market responsible for arranging those resources in advance hasn't been able to meet its own reliability requirement, even at the maximum permitted auction price. PJM must now rely on additional procurement, changes in demand, generating upgrades or other arrangements to close the gap. Those responses may succeed, but they also require time, regulatory approval, financing and physical infrastructure that isn't yet available.

The October emergency brought a different set of constraints into immediate view. In its October 11 application to the Department of Energy, PJM anticipated electricity demand of approximately 111,500 megawatts or more on October 13, alongside roughly 61,591 megawatts of generation outages during the autumn maintenance season. It identified transmission restrictions limiting the movement of electricity between parts of its territory and neighboring systems, and warned that its operating reserves could become unusually tight. DOE issued Order No. 202-26-50 on October 12, authorizing specified emergency measures through October 16. PJM's subsequent operating notices confirmed that it had identified a reliability need for designated generating resources on October 13.

These measures don't establish that blackouts have occurred, or that every resource authorized under the emergency order has been dispatched. They do establish that the operator is making additional resources available through emergency procedures because the ordinary operating configuration may not provide sufficient flexibility. The distinction is important. An emergency intervention that prevents a service interruption is evidence that a protective mechanism worked. It can also be evidence that the system needed protection from conditions it could no longer comfortably accommodate through ordinary operations. Understanding which of those circumstances is developing requires examining what the intervention absorbed, how often similar interventions are needed and whether the conditions requiring them are improving.

Demand Is Growing Faster Than the System Can Accommodate It

PJM's capacity problems have been developing for several years, but the scale of projected demand growth has changed the planning environment. Data centers, particularly facilities supporting artificial intelligence infrastructure, are becoming increasingly important to the operator's forecasts. Unlike many traditional commercial customers, large data centers can require enormous amounts of electricity concentrated at individual sites, with relatively consistent consumption throughout the day. Some facilities also have demanding reliability requirements because even short interruptions can affect equipment, customer services and contractual obligations.

A single large project can therefore create requirements extending well beyond the amount of electricity it consumes. The grid must have sufficient generating capacity to serve that demand, transmission infrastructure capable of delivering it, and enough reserve capacity to manage equipment failures or unexpected changes in operating conditions. Those requirements become more difficult when several large projects seek connections within the same geographic area, particularly where the existing transmission network was designed around different patterns of electricity consumption.

The pace of development is also uneven. A data center can move from planning to construction and operation on a different timeline than a new generating station or major transmission project. New power plants may require years of permitting, equipment procurement, financing, construction and interconnection work. Transmission projects can involve multiple utilities, jurisdictions and regulatory proceedings. Even when the financing is available, the physical system cannot necessarily expand at the same speed as demand. PJM has been working to accelerate interconnections and obtain additional supply, but its capacity auctions indicate that the resources available within the relevant planning periods haven't caught up with projected requirements.

The July auction also demonstrates why higher prices alone cannot be treated as a sufficient response. The price cap was intended to moderate volatility and protect customers from abrupt increases in capacity costs. It provides a degree of financial predictability, which has value for households, businesses and electricity providers. But the cap doesn't create generating capacity. If the resources required to satisfy projected demand cannot be developed or committed within the relevant period, a higher market price may encourage investment without producing enough available supply in time. PJM itself acknowledged that the price collar could limit volatility without resolving the underlying imbalance between demand and supply.

The system's response has consequently expanded beyond its ordinary capacity auction. PJM has proposed a special Reliability Backstop Procurement process intended to secure additional resources, while pursuing changes to how large electricity customers connect to the grid. Both approaches recognize that the existing arrangements may be insufficient for the amount and timing of demand now being anticipated. They also introduce new questions about cost, authority and responsibility. If additional generation must be procured through long-term commitments, someone must assume the financial obligations. If large customers are allowed to connect before enough new supply becomes available, the operator needs enforceable rules governing what happens when the grid becomes constrained. Neither problem can be resolved solely by identifying the demand or announcing the investment required to serve it.

That was the issue behind FERC's October 9 decision. PJM proposed creating a registry of large electricity loads and an interim service arrangement that would allow certain new customers to connect under special curtailment obligations. FERC approved the registry, which will improve the operator's ability to identify and plan for large loads, but rejected the proposed interim service arrangement. The commission raised questions about jurisdiction and whether the proposal adequately justified treating some customers differently from others. Retail electricity service is generally subject to state regulatory authority, while FERC oversees interstate transmission and wholesale electricity markets. PJM's proposal had to operate within those divisions of responsibility.

The rejection doesn't mean regulators have decided that data centers must receive unrestricted electricity service regardless of system conditions. PJM can revise its proposal, and other arrangements may ultimately be approved. FERC's concerns also reflect legitimate questions about the authority to curtail service and the consequences of applying different rules to different customers. A large industrial facility, a hospital, a data center and a manufacturing plant may all consume substantial amounts of electricity, but they don't necessarily have the same ability to interrupt operations or substitute other sources of power. Any enforceable system must account for those differences without arbitrarily transferring obligations among customers.

For PJM, however, the practical consequence is that a proposed mechanism for accommodating additional demand while preserving emergency flexibility remains unresolved. The registry will provide better information about large loads, but information alone doesn't create capacity or establish when consumption can be reduced. The operator can identify a growing constraint without yet having an approved arrangement to manage every aspect of it. Meanwhile, projects continue moving through planning, financing and development, and the electricity system must prepare for the demand they may eventually create.

The Emergency Measures Are Drawing on the Same Resources That Provide Redundancy

The October emergency illustrates how the system's remaining flexibility can become increasingly important even when the immediate cause isn't the same as the longer-term capacity problem. Autumn is normally an opportunity for electricity generators to perform maintenance because demand is often lower than during the summer cooling season or winter heating season. Taking generating units offline during these periods allows operators to repair equipment, perform inspections and prepare for more demanding conditions. The practice is itself a reliability measure. Equipment that isn't maintained properly may become more likely to fail when it is needed most.

PJM's October 11 application identified unusually extensive generation outages during a period when electricity demand was expected to remain elevated. It also described transmission constraints affecting the movement of electricity between different parts of its territory and neighboring systems. Those constraints are important because electricity available somewhere in a regional network isn't necessarily deliverable wherever it is needed. Transmission lines have physical operating limits, and congestion can prevent generating resources in one area from compensating for shortages elsewhere. PJM also identified limitations involving neighboring systems, including the Midcontinent Independent System Operator, which had issued its own capacity advisory and conservative-operations notice.

Interconnection between regional grids is one of the electricity system's most important sources of redundancy. It allows operators to exchange electricity, share available resources and respond to localized shortages without requiring every region to maintain enough generating capacity to meet every conceivable circumstance independently. But that flexibility depends on conditions outside the affected operator's control. Neighboring systems must have electricity available to export, and the transmission infrastructure connecting them must have sufficient capacity to move it. If multiple regions experience elevated demand, maintenance outages or transmission restrictions at the same time, the amount of assistance available to any one system can become limited.

The Department of Energy's October 12 order provided PJM with additional options for managing those constraints. It authorized the dispatch of specified generating resources when necessary to preserve reliability and permitted the operator, under narrowly defined emergency conditions, to direct backup generation at large electricity-consuming facilities. The latter provision is especially consequential because it allows resources ordinarily maintained for private continuity to help reduce demand on the wider electricity system. A data center operating its own backup generators may continue supporting its equipment while drawing less electricity from the grid, leaving centrally supplied electricity available for other customers.

This is a useful arrangement, but it also exposes the limits of treating backup capacity as an uncomplicated solution. Backup generators require fuel, maintenance and equipment capable of operating safely. Batteries have finite storage capacity. Some facilities need their backup resources available for their own essential operations, and operating those resources may involve environmental restrictions or other practical constraints. The DOE order doesn't convert privately maintained backup equipment into permanent regional generating capacity. It authorizes a temporary response under specified conditions, including circumstances immediately preceding or during the most severe level of an energy emergency.

PJM's operating notices provide a more precise indication of what was happening on October 13. The operator had issued a Maximum Generation Emergency/Load Management Alert and instructed generators to provide information about how quickly they could return from planned outages. It also stated that certain resources identified under the DOE order were needed for the October 13 operating day, while clarifying that the order didn't automatically require those resources to operate at maximum output. That distinction prevents the emergency from being overstated. The authorization and the identified reliability need are documented; the eventual amount of additional generation actually dispatched, and the effect on operating reserves, require subsequent operating records.

The October order also follows earlier emergency authorizations involving PJM during 2026. Repeated authorization doesn't establish that the system is approaching collapse. Emergency procedures exist precisely because unexpected or difficult conditions arise, and their successful use can prevent much more serious consequences. But recurrence changes the questions worth asking. If the operator increasingly needs extraordinary measures during periods that were previously suitable for routine maintenance, those measures may be compensating for a reduction in ordinary operating flexibility. If maintenance must repeatedly be postponed or recalled, the system may also be exchanging flexibility in one period for additional equipment and scheduling pressure in another.

The available evidence doesn't yet establish that this exchange is producing a measurable deterioration in equipment reliability. It does establish that PJM is relying on several compensatory arrangements simultaneously: returning generators from maintenance, obtaining emergency operating authority, managing transmission constraints and preserving the option to reduce large customers' dependence on grid-supplied electricity. Each arrangement can work as intended while the underlying conditions remain unresolved. The important measure isn't simply whether customers retain service during the emergency, but whether the system can restore its ordinary operating margins afterward without requiring the same interventions again.

New Power Generation Won't Arrive Until 2028

Private investment is beginning to address some of the capacity constraints, including through agreements that connect major technology companies directly with electricity producers. On October 6, Google and Constellation announced a long-term agreement involving more than $4.3 billion in upgrades to existing nuclear generating facilities across PJM's territory. The upgrades are expected to add approximately 890 megawatts of generating capacity, with the first additions scheduled for 2028. A separate agreement covers 2,700 megawatts of existing electricity supply. The distinction matters because extending commercial support for existing generation isn't the same as creating new capacity, although retaining dependable resources can be valuable when the system is already facing a shortage.

The agreement demonstrates that large electricity customers can contribute to the infrastructure required to serve their demand. Long-term purchasing commitments can provide generating companies with the financial certainty needed to justify upgrades that might otherwise be difficult to finance. Improving the output of existing nuclear facilities may also avoid some of the permitting and construction challenges associated with building entirely new power plants. Demand-response commitments can provide additional flexibility during emergencies. These are meaningful responses, particularly where investment produces additional dependable capacity rather than simply reallocating electricity already available to the system.

But the timing remains consequential. PJM is managing emergency conditions in October 2026, while its capacity auctions have identified shortfalls for future delivery years. The first additional generation under the Google-Constellation agreement isn't expected until 2028. The planned 890 megawatts also represent only a portion of the 6,831-megawatt shortfall identified in July, although those figures aren't directly interchangeable without accounting for capacity accreditation and the timing of delivery. Additional projects may close more of the gap, and some existing resources may remain available longer than previously anticipated. Those possibilities should be included in the assessment. They don't change the fact that financial commitments and announced construction plans cannot supply electricity before the infrastructure becomes operational.

There is another dimension to this investment response. On October 13, the International Monetary Fund's Global Financial Stability Report identified growing financial exposure to artificial intelligence infrastructure. Hyperscale technology companies now account for approximately 10% of global investment-grade bond issuance, compared with less than 1% in 2024, according to the IMF findings reported that day. That doesn't establish that AI infrastructure investment is unsustainable. The companies involved include businesses with substantial revenues, existing assets and access to financing. But it does show that the expansion of AI infrastructure is becoming more important to corporate borrowing markets, while the IMF has identified the potential for weaker-than-expected returns or changes in investment expectations to affect broader financial conditions.

This creates a relationship between two systems operating on different timelines. Electricity providers need long-term commitments to justify infrastructure investment, while technology companies need reliable electricity to support their planned facilities. Financing arrangements can help both parties proceed, but the resulting commitments may remain exposed to changes in construction costs, interest rates, electricity availability and expected demand. A delay in transmission infrastructure can affect when a data center becomes operational. A change in the data center's development schedule can affect the anticipated demand supporting an electricity investment. Changes in financing conditions can influence both projects, even when the underlying need for additional electricity remains.

The allocation of those risks matters. Some contracts place substantial obligations on the companies purchasing electricity. Other projects may involve regulated utility investments, shared transmission infrastructure or costs recovered from broader groups of customers. It would be inaccurate to assume that private investment automatically transfers costs to the public, just as it would be inaccurate to assume that private financing eliminates public exposure. The relevant questions concern who is obligated to pay, what happens if projected demand doesn't materialize, and whether infrastructure built for concentrated private demand creates costs or benefits extending to other electricity users.

This is where the electricity problem begins interacting with a broader financial one. The same investment cycle generating additional electricity demand is also helping finance the infrastructure intended to satisfy that demand. If the cycle proceeds as anticipated, new generating capacity and transmission improvements may reduce pressure on the grid. If demand, financing or construction schedules diverge, the consequences could extend beyond individual technology projects. That is a conditional exposure, not an established failure. The IMF's findings make it more important to examine because the financial commitments involved are becoming large enough to matter outside the technology sector.

Emergency Measures Are Keeping the Grid Operational

The strongest evidence against an immediate reliability crisis is that PJM continues to have functioning mechanisms for managing difficult operating conditions. Its capacity auctions have attracted additional resources, even though they haven't met the full reliability requirement. Emergency procedures allow the operator to recall generating capacity, coordinate transmission and obtain additional operating authority. Federal regulators have approved the large-load registry, improving the information available for planning, while leaving open the possibility of a revised curtailment framework. Private investment is supporting generating upgrades and longer-term supply commitments. None of these developments should be discounted simply because the underlying problem remains unresolved.

The concern is that the mechanisms aren't interchangeable, and success in one doesn't necessarily repair a constraint in another. A backup generator can reduce demand on the regional grid during an emergency, but it doesn't automatically become a permanent source of accredited capacity. A transmission improvement can make existing electricity more deliverable without increasing the total amount generated. A long-term purchasing agreement can support new investment without ensuring that the resulting capacity will be operational during the next emergency. A regulatory decision can preserve appropriate legal authority while leaving an immediate planning problem unresolved.

PJM's recent experience therefore needs to be evaluated across several measures rather than through a single outcome. The capacity auction measures whether enough dependable resources have been committed for future years. Emergency operating notices show how the system responds when available resources become constrained. Regulatory proceedings determine which obligations can be imposed on different participants. Infrastructure investment indicates whether additional capacity is being developed, but its eventual contribution depends on completion and operation. Financial conditions influence whether the investment needed to close the gap remains viable.

Those measures are now interacting more visibly. The July auction identified a substantial future shortfall. PJM pursued additional procurement and a special arrangement for large loads. FERC rejected central elements of that arrangement in October, while an immediate operating emergency required separate federal authorization. At the same time, technology companies are making major electricity commitments, and international financial regulators are identifying growing exposure to the investment cycle supporting their expansion.

The evidence doesn't establish that these developments will produce widespread electricity shortages or a financial disruption. It establishes that several mechanisms responsible for maintaining reliability, accommodating growth and financing additional infrastructure are under pressure at the same time. Some are successfully absorbing that pressure. Others remain incomplete, delayed or subject to regulatory resolution.

The distinction between absorption and restored capacity is central to the assessment. PJM can successfully manage the October emergency without having resolved its 2028–2029 capacity shortfall. It can secure additional generation commitments without eliminating transmission constraints. It can obtain emergency authorization without establishing a permanent framework for managing large loads. Each successful response preserves service or creates additional options, but whether those responses collectively restore sufficient flexibility remains an open question.

The immediate evidence to examine will be the actual operating outcome of October 13–16, followed by the pace at which PJM resolves its procurement and large-load arrangements. A return to ordinary operating conditions, accompanied by increasing dependable supply and more manageable demand forecasts, would weaken the concern. Repeated reliance on emergency measures, persistent capacity shortfalls and growing dependence on resources that aren't yet operational would strengthen it.

For now, the electricity system is continuing to function through a combination of ordinary market arrangements, emergency authority, private investment and operational adjustments. The unresolved issue is whether those arrangements are creating enough additional capacity to accommodate the demands being placed on them, or whether an increasing share of their flexibility is being consumed simply maintaining the service they already provide.

What I’m watching now

The first test is the outcome of PJM's October 13–16 emergency period. Its subsequent operating records should show whether forecast demand and generation outages materialized, how close the system came to its reserve requirements, and whether additional generating resources or privately maintained backup systems were actually dispatched. An emergency order isn't evidence that every authorized intervention was used. The amount of intervention required, the operating margin remaining afterward and the speed with which normal maintenance schedules resume will provide a clearer measure of the system's condition.

The second is whether PJM can resolve the longer-term capacity problem through ordinary procurement and additional arrangements. Its next capacity auction, proposed reliability backstop and revised large-load rules will show whether the operator is securing enough dependable resources to meet projected demand. The relevant measure isn't simply the number of projects announced or the amount of capacity proposed, but how much accredited capacity becomes available within the delivery periods for which it is needed. Changes in demand forecasts also matter. Delayed or canceled data center projects, improved efficiency or reduced electricity requirements could materially improve the outlook without requiring every projected generating project to be completed.

I'm also watching whether transmission constraints and maintenance pressures become recurring features of operating emergencies. A grid can have substantial generation available overall while still struggling to deliver electricity to particular regions. If neighboring systems experience simultaneous shortages or transmission restrictions, the ability to compensate through imports becomes more limited. Repeated maintenance recalls would warrant attention as well, particularly if they begin affecting equipment availability during later periods of high demand. The October emergency provides a documented set of conditions against which future events can be compared.

Finally, the financing relationship between AI infrastructure and electricity development deserves closer examination. The IMF's findings establish that hyperscale investment is becoming a more significant component of corporate borrowing markets. The next indicators are changes in borrowing costs, capital expenditure commitments, project schedules and contractual arrangements between technology companies and electricity providers. A slowdown in AI investment wouldn't necessarily worsen grid reliability; reduced demand growth could relieve some pressure. But if generating investments have already been financed or costs allocated on the expectation of demand that subsequently changes, the consequences will depend on who carries those obligations. That possibility becomes more important as infrastructure investment and large-load development become increasingly dependent on one another.

Current Risk Assessment · October 13, 2026

Current assessment: Elevated and increasing pressure on electricity-system reliability and planning capacity within PJM. The operator faces a documented 6,831-megawatt shortfall against its 2028–2029 reliability requirement, unresolved regulatory arrangements for certain large electricity loads, and an immediate October operating emergency involving generation outages, elevated demand and transmission constraints. Existing compensatory mechanisms remain functional, but several are being relied upon simultaneously.

Expected if this assessment holds: Continued use of emergency operating measures during periods of constrained supply, additional procurement efforts to address future capacity shortages, and growing reliance on long-term private investment commitments. Reliability and cost pressures may persist where demand growth exceeds the pace of completed generating and transmission infrastructure. The severity of those pressures will depend on actual resource availability, demand growth and the effectiveness of operational responses.

Would weaken the assessment: Completion of the October emergency with adequate operating reserves and limited extraordinary dispatch; restoration of ordinary maintenance schedules; successful procurement of sufficient dependable capacity; approval of workable large-load management arrangements; improved transmission availability; and evidence that new generating resources or lower demand forecasts are closing the projected reliability gap.

Status: Elevated. Multiple mechanisms are absorbing stress, but the evidence doesn't yet establish that they are restoring sufficient capacity or operational flexibility to accommodate projected demand. The October emergency is an immediate test of operating resilience, while future capacity procurement and infrastructure completion will determine whether the longer-term imbalance is improving.

---

Sources

• PJM Interconnection — Capacity Auction Procures 138,318 MW of Generation Resources. July 14, 2026. Primary auction results documenting the 2028–2029 capacity shortfall and auction pricing.

• PJM Interconnection — Capacity Market. Updated 2026. Auction reports, capacity requirements and planning materials.

• PJM Interconnection — The Capacity Auction Is Coming. Here's What We're Doing Now. June 2026. PJM's explanation of capacity pricing, supply constraints and demand growth.

• Federal Energy Regulatory Commission — PJM filings and orders, Docket ER26-3515-000. October 9, 2026. Regulatory proceedings concerning the large-load registry and interim service proposal.

• Reuters — U.S. Panel Approves PJM's Registry of Data Centers, Rejects Plan to Curtail Power to Them. October 12, 2026. Reporting on the FERC decision and implications for large-load management.

• U.S. Department of Energy — PJM Emergency Order No. 202-26-50. October 12, 2026. Primary emergency order and supporting application for the October 13–16 operating period.

• PJM Interconnection — Emergency Procedures and Operating Notices. October 11–13, 2026. Primary operational alerts and notices concerning generation availability and emergency measures.

• U.S. Department of Energy — 2026 Emergency Orders. Updated October 2026. Record of emergency authorizations, including earlier PJM orders.

• Reuters — PJM Capacity Market Woes Spur Activity in Bilateral Power Deals. August 26, 2026. Reporting on alternative procurement, bilateral contracts and generating investment constraints.

• International Monetary Fund — Global Financial Stability Report. October 2026. Assessment of financial vulnerabilities, including growing exposure to AI infrastructure investment.

FAA Outage Reveals Redundancy Failures

Aircraft at Newark Liberty International Airport. A September telecommunications failure at the FAA facility controlling Newark-area traffic became a regional disruption after the system's backup fiber path was also unavailable.

In July 2025, the FAA announced that it had completed a new fiber-optic communications network between New York and the Philadelphia facility responsible for Newark-area air traffic. Two months earlier, telecommunications failures had disrupted operations at the Philadelphia TRACON, where controllers handle aircraft arriving at and departing from Newark. The new network was part of the response. The FAA added three high-bandwidth telecommunications connections between New York and Philadelphia, deployed a temporary backup system, and began establishing a local STARS radar-data hub so Philadelphia would no longer depend entirely on radar information coming from New York. By July, the agency said the new fiber connections were running along two separate communications paths so that equipment could continue operating if one path was disrupted.

On September 21, 2026, a telecommunications circuit at the Philadelphia facility failed. The FAA attempted to move communications onto the second path, only to find that a construction crew in New Jersey had accidentally severed the fiber line. Incoming flights were halted at Newark, JFK, LaGuardia, Philadelphia and Boston, and the disruption spread as traffic was redirected through the system. More than 100 flights were diverted. By the FAA's later accounting, roughly 9,500 flights were delayed or canceled, and the telecommunications problem took about eight hours to resolve.

The scale of the disruption got most of the attention, but I'm more interested in the order in which things happened. The FAA had built a second communications path because it knew a failure on the first shouldn't be able to interrupt air traffic across the region. A failure occurred anyway, and the second path had already been lost to an unrelated event.

A fiber line being cut is ordinary enough. So is a telecommunications circuit failing. Put them together in this order, though, and something happening to a fiber line in New Jersey can stop arrivals hundreds of miles away.

The FAA Had Already Added a Backup Route

Had the second communications path been available, most of us probably would've never known the first circuit had failed. One route would've taken over while the other was repaired, and flights would've continued moving.

From the outside, a system in that condition can look unchanged. Inside it, the situation is different. Before the circuit failed, two routes were available. Afterward, one route would've been carrying the work previously distributed across an arrangement that included both. The flights can look the same either way.

We see this elsewhere when electrical grids call demand-response resources or keep generation available that otherwise wouldn't be needed. Hospitals move patients when they run short of beds. Supply chains reroute goods around ports or transportation problems. Government agencies move work elsewhere when an office loses staff.

These responses are supposed to make problems disappear from ordinary life. That's why they exist. It also means the absence of a visible disruption doesn't tell us how much had to happen underneath it to keep things looking normal.

The FAA's response to the Newark problems in 2025 followed the same pattern. The agency wasn't waiting for the next outage. It was adding connections, moving to fiber, building a local radar-data hub and using a temporary backup while the work was underway. It also said that when staffing or equipment problems reduced what controllers could safely handle, it would slow the rate of arrivals into Newark.

By September 2026, the FAA had encountered a problem that the second communications route couldn't take over because the route itself was unavailable, leaving the agency to reduce the number of aircraft entering the affected airspace until controllers had what they needed to handle them safely.

When the second communications route was unavailable, the FAA had to respond outside the telecommunications network by reducing the number of aircraft moving through the affected airspace.

The FAA Is Still Replacing Its Aging Communications Network

The September outage occurred in the middle of a much larger FAA modernization effort. Administrator Bryan Bedford told Congress this month that the National Airspace System is still operating with obsolete technology even as the agency replaces it. As of September 1, the FAA had replaced more than 64 percent of its copper-based Time-Division Multiplexing telecommunications infrastructure with modern equipment and converted more than 400 radio sites. Congress has already provided $12.5 billion for the broader modernization effort.

After the September 21 outage, Transportation Secretary Sean Duffy said the replacement network is supposed to eliminate single and dual points of failure by providing multiple telecommunications routes into FAA facilities. The estimated cost of the telecommunications work has risen to $5.91 billion, and Bedford said the remaining outdated copper telecommunications wiring should be replaced by September 2027.

The FAA is already responding to the weaknesses exposed by the 2025 failures. The difficulty is that recognizing a problem and finishing the work required to fix it happen on very different timelines.

The National Airspace System can't be taken offline while its communications network is replaced. The FAA provides air-traffic services for more than 44,000 flights carrying roughly three million passengers every day, so old equipment has to keep working while new equipment is installed around it. Copper is still carrying traffic while fiber replaces it. Temporary systems have to cover gaps while permanent systems are installed. New equipment has to communicate with parts of the network that haven't been replaced yet.

That leaves a period in which the FAA can know exactly what needs to be repaired, be actively repairing it, and still remain dependent on some of the same equipment and connections that made the repair necessary in the first place.

The September outage happened during that period, when one telecommunications circuit failed while a separate fiber line had already been cut during construction work. There's no indication that one failure caused the other. Their connection was created by the system itself because the fiber line became consequential once the first circuit was gone.

A second route only protects the first route when it's still there at the moment the first one fails. That sounds obvious, but I think it gets lost when redundancy is treated as something a system either has or doesn't have. Its condition can change before anything visible happens.

A New GAO Report Identified Other Communications Vulnerabilities

On the same day as the outage, the Government Accountability Office released a report examining a different set of problems in aviation communications.

GAO looked at electromagnetic-spectrum threats including spoofing and jamming. The FAA has identified those threats, but GAO found that the agency hasn't completed all of the risk and mitigation assessments and updated security documentation needed to address them. The FAA also lacks defined real-time monitoring and detection for every spectrum-related threat GAO examined. In some cases, the agency can investigate an incident only after someone reports it.

GAO made nine recommendations covering risk assessments, monitoring and detection, coordination with other organizations, authentication and protection of aviation communications. The Department of Transportation agreed with all nine, and they remain open.

Nothing in the available evidence connects those findings to the September 21 outage. The fiber damage was reported as accidental construction damage. The GAO report is useful here for what it shows about the communications system around the fiber itself.

For a controller to communicate safely with an aircraft, the FAA has to know where the aircraft is, the controller has to be able to reach it, the information moving between them has to be accurate, and the equipment carrying that information has to work. The FAA also has to be able to recognize when one of those things has changed.

Adding another fiber route protects one part of that chain. It doesn't protect all of it. Another route won't tell the FAA that information has been spoofed, just as better authentication won't restore a cable that's been severed. A second facility also provides less protection if it depends on something that can be lost along with the first one.

Two systems can look independent on paper and still share something underneath them that makes them vulnerable at the same time.

The FAA happened to begin limited use of another new system on September 21. SMART brings roughly 200 data streams into one place, including weather, flight paths, traffic flow and controller staffing. It's intended to help FAA personnel see problems developing earlier and make decisions about traffic before congestion or weather produces larger disruptions.

SMART can give the FAA more information about what's happening across the airspace. The fiber outage gives us an unusually clear example of what information alone can't solve. Officials can know exactly where traffic is accumulating and exactly where they'd like aircraft to go while still lacking the communications route required to move the information controllers need.

The FAA is replacing the network carrying information at the same time it's improving the information available to the people managing traffic. It's also replacing old equipment and trying to close the security gaps GAO identified. Those projects overlap, but they don't solve the same problem, and progress in one doesn't necessarily tell us much about the condition of another.

The FAA Prevented a Safety Crisis by Reducing Air Traffic

When the telecommunications system couldn't support normal operations on September 21, the FAA reduced the amount of traffic controllers had to handle. Arrivals were halted or restricted, aircraft were held on the ground and more than 100 flights were diverted while the telecommunications lines were repaired.

Thousands of passengers experienced delays or cancellations, aircraft were diverted, crews and airlines had to adjust, and other airports received traffic they otherwise wouldn't have handled. Boston was affected even though the telecommunications failure wasn't there because the disruption altered traffic across the Northeast.

This is one reason I've become increasingly wary of describing large systems as either working or failing. The telecommunications network couldn't preserve normal service, while the air-traffic system could still preserve safe service by accepting fewer flights. What failed in one place was accommodated somewhere else, and the cost changed form as it moved.

The FAA reduced traffic until the communications lines could be repaired, shifting the immediate consequences into delays, cancellations and diversions while keeping the affected airspace operating safely.

As the new telecommunications network comes online, I'd expect more of these failures to be handled before they ever reach flight operations. A circuit fails, another route takes over, the first gets repaired and passengers never know it happened. If relatively ordinary technical problems continue making it through the backup systems and forcing the FAA to restrict traffic across several major airports, then we know those backups still aren't containing enough of the problem before it reaches everything connected to them.

What I’m watching now

The investigation into the New Jersey fiber cut should tell us more about how separate the communications routes really were. The FAA's July 2025 announcement described two separate paths intended to keep equipment operating if one was disrupted. I want to know how those routes were physically laid out, what could disable each one and what protection existed around a line that became critical as soon as another circuit failed.

I'm also watching what the FAA says about the temporary backup deployed after the 2025 Newark failures. The agency described it as providing redundancy during the move to the new fiber network, but the public information available so far doesn't establish what it could've carried under the conditions that developed on September 21. Until we know that, I don't think we can say whether another backup was present and unable to do the job, had already been removed from the relevant configuration, or was never intended to cover this particular failure.

More than 64 percent of the FAA's old copper-based TDM telecommunications network had been replaced as of September 1, and Bedford says the remaining outdated copper should be gone by September 2027. The plans described after this outage go further than replacing copper with fiber and call for multiple routes into facilities so that losing one or two doesn't recreate what happened this week.

We should eventually be able to see whether that worked by how uneventful future failures become. If a circuit fails, another independent route takes over and flights continue while the damaged line is repaired, the new network is doing what the FAA says it's being built to do.

The GAO report gives us another set of things we can follow without trying to infer them from outages. All nine recommendations remain open. We can watch whether the FAA completes the missing risk assessments, improves real-time detection of spectrum threats and strengthens authentication and data protection for aviation communications. Those changes would address problems the fiber replacement can't.

Another useful measure will be how often technical problems require the FAA to reduce traffic across multiple airports. Ground stops and traffic restrictions are supposed to protect safety, so their use alone doesn't tell us very much. The more useful question is what keeps requiring them. If relatively ordinary equipment and telecommunications problems repeatedly get past the technical backups and have to be dealt with by moving or stopping flights, then more of the system is depending on that last layer of response than the visible operation of the network would suggest.

Current Risk Assessment · September 25, 2026

Current assessment: The September Northeast aviation outage adds another case in which the existence of a backup tells us less than whether that backup remains available when the original system actually needs it. The FAA had already added separate communications paths after the 2025 Newark failures, but a telecommunications circuit failed after another fiber route had been severed, leaving the agency to reduce traffic across multiple airports while the lines were repaired. The FAA is already replacing the older network, and its response kept aircraft operating safely. The open question is how often problems can still get through the telecommunications backups and reach flight operations while that replacement continues.

Expected if this assessment holds: Individual equipment and telecommunications failures will usually be handled without creating a safety problem, but some will still produce large disruptions when another route or system is already unavailable. As more of the new network comes online, those events should become less common and more failures should be handled without large ground stops, diversions or cancellations.

Would weaken the assessment: Future telecommunications failures move onto independent backup routes without major traffic restrictions; the investigation identifies a specific problem behind the September outage that's subsequently corrected; multiple routes remain available when failures overlap; and the separate communications problems identified by GAO are addressed.

Status: ELEVATED — the FAA is replacing the network, but the September outage showed that two separate telecommunications problems can still reach flight operations before that work is complete.

---

Sources

Federal Aviation Administration, “Secretary Duffy Announces Completion of Another Major Milestone for Operations at Newark Liberty Airport,” July 3, 2025.

Federal Aviation Administration, “FAA Statements on Newark Liberty International Airport,” 2025.

Federal Aviation Administration, “Testimony of Bryan Bedford, FAA Administrator,” September 2026.

U.S. Government Accountability Office, “Aviation Cybersecurity: Enhanced Air Safety Requires FAA to Better Mitigate Threats to Aircraft Communications,” September 21, 2026.

Federal Aviation Administration, “Trump’s Transportation Secretary Sean P. Duffy Delivers State-of-the-Art Air Traffic Control,” September 21, 2026.

Reuters, “Thousands of US flights snarled before FAA resolves telecom problem,” September 21, 2026.

Reuters, “Airport operations in US Northeast resume after telecom lines repaired,” September 21, 2026.

Reuters, “US says new air traffic control system will not suffer disruptions,” September 22, 2026.