Firecracker hosts: $179 a month on Vultr, $300 on AWS, $474 on Azure

Which clouds support nested virtualization, and what a small setup costs on each.

1 view

Plori

Firecracker needs KVM, so the host must be a bare-metal server or a VM with nested virtualization. We priced the same small setup on six clouds, at list prices from October 14 and 15, 2026.

The short answer

  • Vultr costs the least of the providers with an HA control plane: $179 a month with one host.
  • AWS $300, Google Cloud $313, Azure $474. Their $73 control plane fee, higher host prices and egress make the difference.
  • Bare metal costs less once more than 1.7 hosts run all month.

One host, all month, US West

Provider Per month
OVHcloud* $145
Vultr $179
DigitalOcean $196
AWS $300
Google Cloud $313
Azure $474

*No HA control plane, and OVHcloud warns that a live migration can make a nested VM's kernel panic.

When the answer changes

  1. More than 1.7 hosts all month: one OVHcloud RISE-2 bare-metal server costs less, if one failure domain is acceptable.
  2. Hosts that run most of the year: one-year commitments bring AWS, Google Cloud and Azure host prices below Vultr. Their $73 control plane fee stays.
  3. Hosts outside the US, credits, or much more storage and egress: these numbers do not apply.

The full post has what each provider states about nested virtualization, the cost breakdown and Geekbench scores.

Read the full post

Subscribe to Plori Engineering

Engineering notes building AI agents in the cloud

Free. Unsubscribe anytime.

Keep reading

All issues →
  1. Scion-style merges: one comparison takes 1.93 s at 100,000 files

    What we took from Anthropic's Scion for Workspaces, and the merge cost we measured before changing anything.

  2. Claude Code handed 0 of 240 unnamed-server prompts to our MCP server, then 21

    What Claude Code shows the model about MCP tools, and what a rewrite of our server text changed in 240 runs.

  3. AI agent files: a local disk, saved to object storage every 60 seconds

    What a crash can lose, how results are accepted, and the start and save times we measured.