Commercial terms
Two numbers: devices and bytes.
You pay per enrolled device per month, and for the image data your fleet actually pulls. Everything else — check-ins, node pre-warming, rollbacks, retries caused by us — is not billable, because charging for those would create exactly the wrong incentives.
Plans
| Pilot | Fleet | Fleet Plus | Private nodes | |
|---|---|---|---|---|
| Per device / month | free | $0.019 | $0.013 | quoted |
| Device range | ≤ 500 | 5,000 – 250,000 | 100,000 – 1,000,000 | any |
| Monthly minimum | — | $95.00 | $1,300.00 | by contract |
| Egress included | 100 GB, hard cap | 2 TB | 25 TB | by contract |
| Image storage included | 5 GB | 500 GB | 5 TB | by contract |
| Delivery nodes | 4, group A only | all 14 | all 14 + region pinning | 14 + your sites |
| Wave rate cap | 2,000 dev/h | 12,000 dev/h | 40,000 dev/h | unmetered |
| History retention | 60 days | 24 months | 24 months | up to 10 years |
| Signed audit export | no | yes | yes | yes |
| Own signing key (HSM) | yes | yes | yes | yes |
| Support | Documentation and community. | Email, one business day. | Email four hours, phone escalation for an active wave. | Named delivery engineer. |
Above 1,000,000 devices the per-device rate is negotiated and lands between $0.006 and $0.009. Private nodes start at $3,100 per site per month plus a one-off installation charge, quoted against the facility.
Egress
Charged on image bytes delivered to devices, by the region group of the node that served them. Node-to-node replication and pre-warming are not charged.
| Region group | Nodes | Fleet | Fleet Plus |
|---|---|---|---|
| A — Europe and North America | 11 nodes | $0.018 / GB | $0.014 / GB |
| B — Asia-Pacific | sin-a, nrt-a | $0.026 / GB | $0.021 / GB |
| C — South America | gru-a | $0.031 / GB | $0.025 / GB |
| Image storage beyond the plan allowance | origin | $0.021 / GB-month | $0.021 / GB-month |
| Cold retention beyond 24 months | origin | $0.008 / GB-month | $0.008 / GB-month |
Storage is counted after deduplication
Blocks shared between builds of the same product line are stored and billed once. A 1.4 GB terminal image whose successor differs by 96 MB adds 96 MB to your storage line, not 1.4 GB. Sixty builds of a 3.2 MB meter image typically occupy under 200 MB.
Two worked examples
Both use the device classes from the overview. Neither is a best case.
Payment terminals — where egress dominates
Fleet38,000 terminals, 1.4 GB image, one full-image release this month because the base OS changed. Served from Ashburn and Dallas — group A.
| Line | Quantity | Amount |
|---|---|---|
| Devices | 38,000 × $0.019 | $722.00 |
| Egress served | 52.0 TB | — |
| Included | −2.0 TB | $0.00 |
| Egress billable, group A | 51,200 GB × $0.018 | $921.60 |
| Image storage | 19.6 GB of 500 GB | $0.00 |
| Monthly total | — | $1,643.60 |
Bytes are 56 % of the bill. Had the same release gone out as deltas — 96 MB per device, 3.48 TB total — the egress line would have been $27.28 and the total $749.28. This is the entire economic argument for delta updates, in one row.
Smart meters — where devices dominate
Fleet Plus480,000 meters, 3.2 MB image, two delta releases this month at 180 KB each. Served from Frankfurt, Warsaw and Madrid — group A.
| Line | Quantity | Amount |
|---|---|---|
| Devices | 480,000 × $0.013 | $6,240.00 |
| Egress served | 164.8 GB | — |
| Included | 25 TB | $0.00 |
| Egress billable | 0 GB | $0.00 |
| Image storage | 192 MB of 5 TB | $0.00 |
| Monthly total | — | $6,240.00 |
Even a full-image release to the whole fleet — 1.46 TB — would stay inside the 25 TB allowance. For small images the device line is the bill, and the only lever that matters is the per-device rate.
What is not billable
Check-ins
About 1.6 MB per device per month at the default interval. It is control traffic, it is not image data, and it is not metered — otherwise you would be tempted to check in less often, which is the opposite of what a safe fleet does.
Pre-warming and replication
Getting a release onto all 14 nodes before the canary starts costs us bandwidth and costs you nothing. Under 2 % of network egress, and it is the reason the first cohort is not slow.
Our own retries
A block re-fetched because a node served it badly is not billed. A transfer resumed after the device lost power is billed only for the bytes that actually crossed.
Rollbacks
A rollback moves no bytes — the previous image is already in the other slot. You are never charged for an update that failed.
Halted waves
You pay for what was delivered before the halt and nothing after. A release that stops in the canary cohort costs the bytes 480 devices pulled.
Dormant devices
A device that does not check in at all during a month is not counted. Devices in a warehouse, or seasonal fleets, cost nothing while they are dark.
Service commitments
| Commitment | Target | Measured as | Credit |
|---|---|---|---|
| Control-plane availability | 99.9 % / month | Successful API responses over total, excluding 4xx. | 10 % of the monthly fee per 0.1 % below target, to 50 %. |
| Delivery availability | 99.95 % / month | Device-observed: transfers that completed against transfers attempted, network-wide. | 10 % per 0.05 % below target, to 50 %. |
| Halt actuation | 60 seconds | From the threshold being crossed to no further devices being offered the build. | 10 % of the monthly fee per incident. |
| Scheduled maintenance | ≤ 4 h / month | Announced at least 5 days ahead, published on Status. | Excluded from availability. |
Not covered: failures caused by a customer image, device-side faults, or a fleet's own network. We will help you diagnose all three, and we will not pay credits for them. Full terms in Legal.
Start a pilot
Tell us about the fleet
A delivery engineer replies within one business day with a sizing estimate and a flash-budget assessment for your target. No account is created and nothing is charged.