Ninety-seven percent uptime sounds like a demanding promise. It is less demanding than it sounds, and the clause around the number matters more than the number itself.
The federal NEVI rule, 23 CFR 680.116(b), requires each charging port to average above 97 percent uptime over a year. Take a 525,600-minute year. Three percent of that is 15,768 minutes, or about 10.95 days. A port could sit dead for nearly eleven days of unexcluded downtime and still clear the bar. If the dead port is one of four at your site, drivers who needed it that week will not describe your site as 97 percent reliable.
So when a vendor or a funder says "97 percent," ask the questions that give the number meaning. This article covers the host-agreement clause only. For the wider list of things to require in a bid, see the RFP checklist, and for revenue share and exit terms, see charging-as-a-service contract terms.
Where the 97 percent comes from, and who it binds
The 97 percent standard is a federal requirement on charging projects funded under the NEVI Formula Program and other Title 23 funds covered by 23 CFR Part 680, enforced through the state or other direct recipient. It does not apply to a privately funded site unless your contract says so. A vendor offering 95 percent or 99 percent is not breaking a rule either way. If your project is funded through NEVI or a state program that borrows the standard, the funding agreement carries the obligation and your host agreement should match it. Confirm with the funding agency whether your award is covered; see federal programs and the NEVI second-round overview for the funding side.
State programs write their own versions. California's Energy Commission, for example, has published reliability terms with their own exclusions, summarized in public reporting as covering vandalism (capped per incident), grid power loss, natural disasters, and a capped allowance for planned maintenance. Those are program terms, not the federal text, and they differ from it. Copy the definition your funder uses, not one remembered from another program.
How uptime is measured under the federal rule
The rule gives a template you can borrow almost line for line:
- Per port, not per site. A site with four ports and one dead all month still has three working, and a site-level average hides the dead one.
- Hardware and software both count. A port is up when it is online and available, or in use, and it actually delivers power at the required minimum level. A unit that shows "available" in the app but will not start a session is down for drivers, and should be down in the contract.
- A monthly calculation over the previous twelve months (as the regulation text is described in the sources named at the end; check paragraph (b) yourself before copying it). The yearly figure is the one that counts, so a bad month can be absorbed by good ones. Your contract can be stricter and test every month.
- Outage minutes net of excluded minutes. Uptime is 525,600 minutes minus outage minutes (less excluded minutes), divided by 525,600. The excluded list in the federal rule covers events outside the operator's control: electric utility service interruptions, a failure to charge caused by the vehicle, scheduled maintenance, vandalism, and natural disasters, plus hours outside the station's stated operating hours. The operator has to be able to show the port would otherwise have worked.
Working the monthly figure the same way: a month is roughly 43,800 minutes, so 3 percent is about 1,314 minutes, a little under 22 hours per port. Two long outages can use most of it.
Clause structure that does real work
Every host agreement is different, and this is a structure to discuss with counsel, not a form to sign. These are the pieces worth having.
1. Definition. Copy the federal wording or your funder's: hardware and software online, available, and delivering power. State that a port that cannot complete a paid session is down, whatever its status light says. If you are on a payment-heavy site, say that a payment terminal fault that blocks all sessions counts too.
2. Measurement. Per port, calculated from the network's own logs, reported monthly. Name the data source, the reporting date, and what you receive: port-level outage start and end times with error codes, not a single summary percentage. Reserve the right to audit the logs.
3. Target. A number, per port, over a stated period. Choose the period deliberately. An annual average tolerates a ten-day outage; a monthly test with a cap on consecutive days does not. If you adopt 97 percent, adopt the same measurement basis, or the number means something different in your contract than in the rule.
4. Exclusions, each capped. An open-ended exclusion list is where an SLA goes to die. For each carve-out, say who bears the burden of proof, and set a cap where it makes sense: planned maintenance with a minimum notice period and a maximum number of hours per year, vandalism excluded only if reported and repaired within a stated window, utility outages excluded only if confirmed by the utility. Decide whether you will accept "vehicle fault" or "driver misuse" exclusions at all.
5. Repair obligation. Response time and repair time, separate from the uptime number. A target on its own can be met by one lucky quarter; a deadline to dispatch a technician and fix a dead port within a set number of days addresses the actual failure.
6. Remedy. The usual structures are service credits against fees or revenue share, a stepped schedule (a larger credit per percentage point missed, per port, per month), an accumulating right to terminate after repeated misses, and a right to repair at the vendor's cost if the deadline passes. We could not find a published benchmark for how large the credit or penalty should be, so no figure is offered here. Size it against what a dead port costs you in lost revenue and tenant goodwill, and ask each bidder what they will accept.
7. Liquidated damages, if used. A fixed payment for a missed target is called liquidated damages. As a general matter courts tend to enforce such a clause when it is a reasonable forecast of loss that is hard to calculate, and to strike one that works as a punishment, and the standard differs by state. That is a reason to tie the amount to a stated rationale such as lost revenue per port-day and have counsel review it, rather than choosing a round number. A public-sector example exists: a 2026 Massachusetts state service-plaza solicitation was reported to pair a 97 percent charger uptime requirement with liquidated damages under a lease appendix. We could not read the document itself, so treat that as a pointer to look for, not a model to copy.
8. Reporting and exit. Monthly reports, a notice duty when a port is down more than a set number of days, and a termination right if the vendor misses the target in a defined number of months in a row. If you are in a Charging-as-a-Service arrangement, tie the exit right to the buyout formula so that underperformance does not leave you locked in with equipment you do not own.
Mistakes to avoid
- Letting the vendor define the data. If the only uptime number you ever see is the one the vendor reports, you cannot check it. Require port-level logs.
- Accepting a site-level average. One dead port in four drops a site's average by only a quarter of that port's downtime.
- Excluding "network outages." If the vendor's platform is down and your chargers cannot take payment, drivers cannot charge. That is the vendor's failure and should count.
- Confusing the SLA with the warranty. A hardware warranty covers parts and sometimes labor. An uptime SLA covers outcomes. You usually need both, and the maintenance cost budget has to pay for whichever one is not in the vendor's price.
A short pre-signing test
Before you sign, run one scenario against the draft: a single port fails on a Friday and the part ships Tuesday. Is it down? Is the weekend excluded? When does the clock start, who has to report it, and what happens to the fees that month? If the contract cannot answer in a sentence or two, the clause is not finished.
This article is general information, not legal advice. Have counsel experienced in commercial property and service agreements review any uptime clause before you sign.
Last factually verified: November 2, 2026. The 97 percent per-port standard, the 525,600-minute formula, and the exclusion for hours outside operating hours were re-checked on November 2, 2026 against the Cornell Legal Information Institute copy of 23 CFR 680.116 as returned by search. The full list of excluded events and the definition of an "up" port are carried from our RFP checklist, which read 23 CFR Part 680 directly at ecfr.gov on November 1, 2026; the ecfr.gov, Cornell, and other primary hosts were not reachable from the environment that produced this article, so we did not re-read the exclusion list ourselves. The California reliability terms are as summarized in search results for the California Energy Commission's published agreement and were not read at the source. The Massachusetts solicitation was identified from a search listing only. We did not find a published benchmark for penalty size or a sample host agreement, so none is given. evcharginghelp.com has no relationship with any vendor or program mentioned.
evcharginghelp.com is editorially independent and receives no compensation from any company mentioned.