All systems operational 14 regions 99.998% uptime (90d)
v3.4.2 shipped Apr 24 — read changelog
Compute

Dedicated cores.
Real ones.

Every vCPU on Arveon is a thread on a hyperthreaded core that nothing else is using. NVMe-only storage. AMD EPYC Genoa or ARM Ampere Altra. Boot in under twelve seconds, billed by the second.

From €0.022/hour·14 regions·x86_64 & aarch64
Plan a8 · eu-hel-1 running
v0
v1
v2
v3
v4
v5
v6
v7
8 dedicated vCPU · 16 GB ECC · 320 GB NVMe

CPU steal0.00%
NVMe IOPS (4k r)412 000
Read latency, p9994 µs
Inter-AZ rtt0.42 ms
Built for production

No noisy neighbors. No burst credits. No asterisks.

Dedicated vCPUs

A vCPU is a hyperthread on a physical core that nothing else can touch. CPU steal is exactly 0.00% — measured continuously and exposed in the console.

NVMe-only storage

No spinning disks. No "general purpose SSD with bursting." 412 000 random-read IOPS at p99 94 µs, on every plan from a2 up.

12-second provisioning

Median p50 is 11.4s, p99 under 22s. We measure it every 60 seconds in every region and post the numbers on our status page.

Per-second billing

Boot at 13:42:17, destroy at 13:43:08, pay for 51 seconds. No minimum charges. No "rounded to the hour."

Encrypted at rest

LUKS2 with per-volume keys, sealed in our HSM. Keys are rotated on snapshot, never leave the region, and you can supply your own (BYOK).

Live migration

When we patch hypervisors, your VM moves to fresh silicon in under 800 ms. No reboot, no scheduled maintenance windows, no email.

Plans

Pick the size, not the architecture.

Each plan ships with the same NVMe class, the same network throughput per vCPU, and the same SLA. Scale by changing one flag.

Plan vCPU RAM NVMe Transfer € / hour € / month
a2 Hatch24 GB80 GB4 TB€0.022€16
a448 GB160 GB8 TB€0.058€42
a8 Forge816 GB320 GB16 TB€0.119€86
a161632 GB640 GBunlimited€0.236€172
a24 Anvil2464 GB1.2 TBunlimited€0.583€420
a4848192 GB3.2 TBunlimited€1.722€1 240
a96 Bare metal96512 GB7.6 TBunlimited€4.080€2 940

aarch64 plans are billed at 78% of the listed price. Sustained-use discounts apply automatically after 720 hours.

In code

Provision nine nodes across three AZs.

$ arv compute create \
    --plan a8 --image ubuntu-24.04 \
    --region eu-hel-1 \
    --az "1a,1b,1c" --count 9 \
    --tag tier=api --tag env=prod \
    --user-data @cloud-init.yaml

# 9 nodes ready in 14.2s
api-eu-hel-1a-001  185.4.118.92   running
api-eu-hel-1a-002  185.4.118.93   running
api-eu-hel-1b-001  185.4.119.41   running
api-eu-hel-1b-002  185.4.119.42   running
api-eu-hel-1c-001  185.4.120.18   running
resource "arveon_compute" "api" {
  for_each = toset(["1a", "1b", "1c"])

  name   = "api-eu-hel-${each.key}"
  image  = "ubuntu-24.04"
  plan   = "a8"
  region = "eu-hel-1"
  az     = each.key
  count  = 3

  tags = { tier = "api", env = "prod" }
}
Customer story · Tidewave
"We migrated 12 services off a hyperscaler over a long weekend. The savings on egress alone paid back the entire migration in six weeks."
EL
Elin Lindqvist
VP Infrastructure · Tidewave
62%
Lower monthly bill
0
Egress fees in 14 months
14.2s
Mean fleet provision
4
Engineers on the migration
Related

What pairs with Compute.

Spin up a node. See for yourself.