Market-observed fair ranges.
Not a price list.
Node providers set their own prices; the real numbers appear in the wizard as you compose an engine. The tables below summarise what the market currently charges — so you can estimate budget, sanity-check a quote, or plan an engine before you start clicking. Prices vary by node class, region, hosting model, and privacy features. Premiums and discounts stack.
Three independent axes of variation.
Bigger class → more vCPU, RAM, SSD per node → higher monthly. Step-ladder across five classes; see the canonical catalog. Class is the single biggest driver of price.
A performance node in Zurich can cost 40% more than the same node in N. Virginia. Switzerland, the UAE, and São Paulo are the premium tiers; Hetzner EU is the discount floor; Tokyo and Singapore sit above EU hyperscalers.
SEV-SNP, dedicated bare-metal, and Gen II virtualized slices each attract a measurable premium. Renewable energy and compliance attestations typically don't — providers recover those through reputation.
By class, by region.
USD per node per month, covering infrastructure ownership only. Bandwidth and any metered compute are additional — see below. Ranges span the typical spread between cloud-backed and thinly-marked-up dedicated offerings within each region.
| Class | Spec floor | United States | EU (hyperscaler) | APAC | LATAM | Hetzner EU |
|---|---|---|---|---|---|---|
nano* · Nano | 2 vCPU · 4 GiB | $32–$33 / mo | $33–$36 / mo | $38 / mo | not in region | not in region |
std · Standard | 8 vCPU · 32 GiB | not in region | $280 / mo | not in region | $320 / mo | $76–$84 / mo |
perf · Performance | 16 vCPU · 64 GiB | $470–$490 / mo | $500–$540 / mo | $580–$590 / mo | not in region | $148–$156 / mo |
hperf · High-Performance | 32 vCPU · 128 GiB | $960 / mo | $1,020–$1,060 / mo | $1,140 / mo | not in region | $300 / mo |
dense · Dense | 64 vCPU · 256 GiB | $1,820 / mo | not in region | not in region | not in region | not in region |
How to read a range: the low end reflects thin-markup cloud-backed providers and high-volume regions; the high end reflects provider-operated dedicated hardware or premium-region SKUs. Add a feature premium (below) for SEV-SNP or dedicated bare-metal.
* Provisional class — final capacity allocation pending validation.
What specific features add to a node.
Apply these on top of a class × region base price. Premiums are roughly additive within reason — a dedicated SEV-SNP node in the EU will cost considerably more than a plain cloud-backed node, but both are bounded by the market.
| Feature | Typical premium | Why |
|---|---|---|
Dedicated bare-metal (no SEV) | +40 to +80% | Whole-machine provider-operated hardware over the cloud-instance baseline. Combines hardware ownership (~$370/mo amortised on a Gen II machine), thinner operator volume, and isolation. |
SEV-SNP, dedicated bare-metal | +25 to +40% | AMD confidential-computing root-of-trust, applied on top of the bare-metal premium. Boot attestation plus encrypted memory; runs only on dedicated hardware operators control end-to-end. |
100% renewable energy | ~0% | Facility-level grid contracts pass through at parity — the business case is reputational, not cost-recovery. |
Compliance (SOC 2 · ISO 27001 · ISO 27017) | ~0% | Enterprise-ready compliance overlays don't cost more at the per-node level — operators recover the audit expense through reputation and customer preference. |
The second-largest cost after the node itself.
Egress bandwidth is charged per GB by the node's operator and varies wildly by source. Network-heavy apps should pay close attention to the mix. For reference, a moderately active production app serving 3 TB/month of egress will spend ~$270 at AWS rates, ~$33 at Hetzner rates.
Standard egress on m6i / r6i instances. Free tier applied per instance.
Undersea-cable surcharge on the standard rate.
Within ±1% of the AWS price anchor at the same class.
Tokyo / Singapore APAC surcharge.
No free allowance; sliding rate by destination.
Bundled per instance — for most workloads, effective cost rounds to the allowance.
Colocation / transit contracts run an order of magnitude below hyperscalers.
Four typical engine shapes and their ballpark monthly spend.
- · 4 × nano nodes (2 vCPU · 4 GiB)
- · Mix of Hetzner CCX13 (Falkenstein) and AWS t3.medium
- · Egress: 100 GB included on AWS + $0.09/GB; Hetzner bundles 20 TB
- · 7 × standard nodes (8 vCPU · 32 GiB) from 7 providers
- · Mix of Hetzner CCX33 and hyperscaler m6i.2xlarge in EU
- · Egress: hyperscaler 100 GB included + $0.087–0.09/GB; Hetzner 20 TB bundled
- · 13 × performance nodes (16 vCPU · 64 GiB) spanning US, EU, APAC, LATAM
- · All hyperscaler — m6i.4xlarge, c3-standard-16, D16s v5
- · Egress: $0.087–0.114 / GB across regions
- · 13 × high-performance nodes (32 vCPU · 128 GiB)
- · All EU-resident — GDPR-contractual
- · SEV-SNP on dedicated bare-metal only (cloud SEV-SNP not offered)
- · Egress: dedicated / colocation $0.012–0.025 / GB
Low end of each range assumes budget operators (Hetzner, thin-markup cloud-backed). High end assumes premium regions or dedicated bare-metal with SEV-SNP. Real engines usually sit in the middle.
Everything your canisters do is paid in cycles.
Persistent memory
Storage held inside canisters.
CPU cycles
Update and query computation.
Inter-canister calls
Cross-canister and cross-engine messaging.
Ingress & egress bandwidth
External requests hitting your engine.
Cycles are prepaid. The wizard's payment step splits the up-front charge into two buckets you configure independently: an operating budget (default 4 weeks) and an emergency cycles balance (default 2 weeks). When the operating balance exhausts, the engine freezes — canisters paused, data retained — and the emergency reserve keeps it alive for the window you chose. Top up anytime during that window to resume. Set the emergency reserve to zero only if you actively accept the risk of permanent deletion on cycle exhaustion. Recurring auto-top-up is on the roadmap.
See real prices next to real nodes.
The engine wizard shows every offering's actual monthly ownership, bandwidth, and compute rates from its provider — side-by-side with a live readout of your engine's total and estimated capacity. Use this page to plan, then check the wizard to confirm.