Almost nothing about a signaling upgrade is visible from the platform. There is no ribbon cutting, no new station canopy, no map redesign. A rider who takes the same train for a decade might never learn that the relays governing their commute were replaced by radio-based computers. And yet signaling is the single biggest determinant of how many trains a line can run, how closely they can follow each other, and whether a dispatcher's screen reflects reality. It is the quietest capital program in American transit and arguably the most consequential one. It is also, as two very bad summers of safety-records scandals have reminded us, only one layer of a much larger trust problem.
Two Crashes That Rewrote the Rulebook
Modern train control in the United States is not primarily a story about throughput. It is a story about what happened when the machines stopped seeing trains.
Fort Totten and the Circuit That Went Blind
On June 22, 2009, a Red Line train struck a stopped train outside Fort Totten station on WMATA's system. Nine people died and 52 were injured. The NTSB traced the cause to a failed track-circuit module that let the automatic train control system lose detection of the standing train - the signaling equivalent of a blind spot that the system itself did not know it had. WMATA's response was to pull its operators out of automatic mode and run trains manually. That stopgap lasted roughly fourteen years; full automatic train operation was not restored until 2023. An entire generation of Washington commuters rode a system running below its design capability because the agency no longer trusted its own train detection.
Chatsworth and the Federal Mandate
The commuter-rail equivalent came on September 12, 2008, when a Metrolink train ran a red signal and collided head-on with a Union Pacific freight in Chatsworth, California. Twenty-five people died and more than 135 were injured. Congress passed the Rail Safety Improvement Act within weeks, mandating Positive Train Control on most passenger and hazmat routes by 2015 - a deadline the industry missed, then met after an extension to December 31, 2020. The FRA confirmed full nationwide implementation on December 29, 2020. PTC now governs roughly 59,000 route-miles of American railroad, according to FRA figures from October 2024. Amtrak had already been running an early cousin of it, the Advanced Civil Speed Enforcement System, on the Northeast Corridor since 2002.
The casualty list in between is the argument for the mandate: Spuyten Duyvil in 2013 (four dead), Amtrak 188 in Philadelphia in 2015 (eight dead), DuPont, Washington in 2017 (three dead, on a corridor where PTC was installed but not yet active), and Cayce, South Carolina in 2018 (two dead). Each was classified as preventable by the technology that was already on order.
What CBTC Actually Buys You
On rapid transit, the parallel technology is communications-based train control. Where PTC is fundamentally an enforcement overlay - it stops a train the crew is mishandling - CBTC replaces the signaling system outright, and in doing so changes the physics of capacity.
From Fixed Blocks to Moving Blocks
Legacy signaling divides track into fixed blocks and allows one train per block. The block length is set by worst-case braking distance, which means a line's capacity is frozen at whatever assumptions engineers made decades ago. The New York City subway alone has roughly 14,850 signal blocks and 339,191 relays keeping that logic running. CBTC swaps the fixed grid for continuous two-way radio reporting: each train knows and transmits its own position and speed, and the system computes a safe separation envelope in real time. Trains can run closer together without running less safely.
The L Train Proof of Concept
The Canarsie Line is the demonstration case. The MTA commissioned Siemens Trainguard MT CBTC on the L in 2009 across roughly 17 kilometers of route, and peak throughput went from about 12 trains per hour to 26 - more than doubling capacity on a line whose tunnels and platforms never changed. The 7 line followed. For an agency that cannot easily build new tunnels under Manhattan, that is the cheapest capacity in the portfolio.
BART, Chicago, and the Capacity Math
BART's Train Control Modernization program bundles CBTC with 252 new railcars plus yard and substation work, with Hitachi Rail STS USA awarded the design-build contract on September 30, 2020. The target is to lift peak throughput from 24 to 30 trains per hour by around 2032, delivered across eight phases. Chicago's $2.1 billion Red and Purple Modernization Phase One - the CTA's largest capital project ever - pairs corridor signal and advanced-systems work with the Red-Purple Bypass that untangled a century-old junction, and delivered four new stations in 2025. Capacity gains of 25 percent without new right-of-way are the kind of return that makes transit economists at the Eno Center for Transportation sit up.
The 85 Percent Problem
Here is the pivot that keeps this from being a good-news story. As of 2025, roughly 85 percent of the New York City subway still runs on fixed-block signaling, per The New York Times reporting in April 2025. The L and the 7 are the exception, not the leading edge of a completed wave.
Relays, Delays, and a Falling Cost Curve
Aging relays are not an abstract maintenance concern - they are a service-quality problem. The Times reported in 2017 that signal equipment failures accounted for about 13 percent of subway delays in 2016. The encouraging development is cost. By shifting to design-build procurement, the MTA has cut CBTC installation from roughly $50 million per mile to about $25 million per mile, according to Streetsblog NYC in May 2025. The agency committed to 200 track-miles between 2020 and 2029, and the 2025-2029 capital plan budgets $5.4 billion for 75 miles covering the N/Q/R/W, J/Z, and A/C/E.
Queens Boulevard, Contract by Contract
What that looks like on the ground: Queens Boulevard East carries a $62.65 million Mitsubishi contract awarded in 2022 targeting 2026, alongside a $205.8 million Siemens/Thales package. One corridor, a quarter-billion dollars, most of a decade. Halving the per-mile cost is real progress, and it still means resignaling the system is a multi-decade commitment rather than a capital cycle.
The Money Question
Signal work competes for the same dollars as everything else that is wearing out.
A $105 Billion Backlog
FTA budget documents for FY2022-25 put the national transit state-of-good-repair backlog above $105 billion, spanning more than 27,000 buses, 2,200 rail cars, 200 stations, 300 maintenance facilities, and hundreds of miles of track, guideway, power, and signals. FTA state-of-good-repair formula funding was $4.3 billion in FY2025, and Capital Investment Grants are authorized up to $4.6 billion a year. The arithmetic does not close.
The December 2026 Squeeze
Both pools are now operating under the IIJA stopgap running through December 11, 2026, a deadline we covered in detail in the IIJA advance-appropriations explainer and in our look at the broader transit fiscal cliff. Multi-year signaling programs are uniquely exposed to this kind of uncertainty: they are long, phased, and contractually unforgiving. BART's eight-phase schedule depends on a funding picture that, as we noted in our coverage of the agency's FY27 budget and November ballot strategy, is not yet settled.
What Train Control Can't Automate
It is tempting to frame automation as the answer to human error. It is a partial answer, and being honest about the limits is the point.
Two Scandals, One Pattern
In 2026, two of the country's largest agencies produced the same category of failure. Seven MBTA employees were indicted on July 9 and 10, 2026, for falsifying Red Line track-inspection reports and fabricating train numbers between January and October 2024. Days earlier, reporting from TWU and ABC7 on June 30 and July 1 exposed MTA bus-maintenance managers signing off on brake, steering, and tire repairs under "redline hold" that were never performed. We covered each in depth - the MBTA indictments, the MTA bus-maintenance case, and what the two together say about oversight - so the relevant question here is narrower: would CBTC have caught any of it?
Parallel Systems, Parallel Risks
Partly, and only in one direction. CBTC and PTC replace human-attested block occupancy and train separation with continuous machine-reported position and speed. That structurally eliminates one category of manual-certification failure, because no one signs a form asserting where a train is - the train says so, continuously, and the system acts on it. Notably, the MBTA has no large CBTC program comparable to the MTA's, BART's, or the CTA's; its defining 2024-2026 signal-adjacent story is an inspection-fraud scandal, which is a process failure rather than a technology one.
But track inspection, brake overhauls, tire replacement, and switch maintenance all run parallel to train control, not inside it. A moving-block system will happily compute a perfect braking curve for a train whose brake job exists only on paper. Automation narrows the surface area for falsified paperwork; it does not change the incentives that produce it. Those live in staffing levels, supervision, audit independence, and whether an inspector who reports a defect creates a problem for themselves - the same terrain we explored in rethinking who keeps the subway safe. TransitCenter and the Eno Center have both argued that oversight capacity, not sensor count, is the binding constraint, and the 2026 cases are strong evidence for that view.
Finishing an Upgrade That Has to Be Done Twice
The next decade of American transit capacity will be won mostly in equipment rooms. Seventy-five miles of New York resignaling, BART's run at 30 trains per hour, the CTA's rebuilt North Side corridor, and 59,000 route-miles of PTC that now need sustained maintenance rather than installation - all of it adds riders' worth of capacity without a single new tunnel. That is the best value proposition in the industry, and it deserves to survive whatever Congress does in December.
The caveat is just as durable. Every dollar of signaling investment buys a machine-verified answer to one question - where is the train, and how fast is it going - and no answer at all to another: was the work actually done? The agencies that get the next decade right will treat those as two separate programs, funded separately and audited separately. The silent upgrade is worth finishing. It just is not the whole job.