If the station near you always seems broken, you are probably right. Station failure is not evenly distributed. A specific, identifiable class of stations fails constantly while others almost never do — and the uptime numbers networks publish are calculated in ways that hide the difference.
We measure uptime independently, from real charging sessions across our coverage area. Here is what the measured numbers say, why they disagree with the networks’ claims, and how to tell if your local station is in the always-broken class.
The gap between claimed and measured uptime
Most networks publish an uptime figure. Almost all of them self-report it, define “up” generously, and average it across their whole fleet.
Our session telemetry says the real numbers run lower. EVgo claims 98% uptime. We measure 95.4%. Electrify America claims 97%. We measure 96.0%. ChargePoint publishes no fleet-wide figure at all because host sites maintain their own hardware — we measure 94.8%. Blink makes no public claim. We measure 91.2%, the lowest of any major network. The full comparison is in our network table, under “Uptime (measured vs. claimed).”
A few percentage points sound small. They are not. At 91.2% uptime, roughly one stall in eleven is dead at any given moment. If your neighborhood site has two stalls, there is a meaningful chance at least one is down every time you arrive — and a two-stall site with one dead stall functions like a one-stall site with a queue.
Why the self-reported numbers run high
A stall counts as “up” in most networks’ reporting if it responds to a network ping. That includes stalls that power on but fail the payment handshake. It includes stalls that start a session and die two minutes in. It includes stalls delivering 20 kW on a 150 kW cabinet.
From the driver’s side, all three of those are broken. From the network’s reporting side, all three are up. That definitional gap is most of the difference between a 98% claim and a 95.4% measurement — and our telemetry counts a session as successful only when it actually delivers energy without an error.
The three station classes that are actually always broken
Failure concentrates in three specific classes of station.
Older Level 2 units. First-generation public Level 2 hardware from the early 2010s is the single worst category we measure. Screens dead from sun exposure, connectors with ten years of drops, and no revenue incentive for anyone to fix them. Many are hosted by property owners who stopped paying attention years ago.
Single-stall sites. One stall means no redundancy and no queue data. When it fails, the site’s availability goes to zero, and it stays down longer — a single dead stall at a strip mall does not trigger the response a dead eight-stall highway site does.
High-traffic urban DC fast sites. Heavy cycling wears connectors and cables faster than anything else. An urban site running near-constant sessions accumulates a year of highway-site wear in a few months. This is why the busiest station in your city is so often the flakiest.
If your local station is an older Level 2, a single-stall site, or a hammered urban fast charger, it is not cursed. It is in the failure class.
Who actually fixes a broken stall — and how fast
The other half of “always broken” is repair speed, and it varies enormously by who is accountable for the hardware.
Tesla owns and maintains every Supercharger. Typical fix time in our data is under 48 hours. Electrify America and EVgo operate their own equipment and run 24/7 support lines. Typical fixes take one to two weeks. ChargePoint and Blink sell hardware that hosts operate. When the host is an attentive business, fixes happen. When the host is a parking garage that installed chargers for a 2015 grant, a dead stall can stay dead for months. Nobody at the network is accountable for it.
This ownership difference — not charging technology — is the main reason the same brand of station can be reliable in one location and permanently broken in another.
How to protect yourself
Check recent session activity, not the station’s listed status. An app showing a station as “available” tells you it pinged the network. A station that shows completed sessions in the last few hours actually works.
Prefer sites with four or more stalls. Redundancy is the single best predictor of a successful visit in our data.
Report the failure to the network, not just the app. Host-operated stalls especially — the host often does not know it is down.
And if you are choosing where to charge regularly in Seattle, Portland, or Denver, our city guides flag the sites with strong measured reliability, not just the ones that exist on a map.