[ 00 / hero ]
aes services // v1.0 — public launch · june 15, 2026

Raw metal.
No nonsense.
No cloud tax.

Metalhost is bare-metal & VM hosting for developers and startups who'd rather read a datasheet than a marketing deck. Real cores. Real disks. Transparent pricing. Provisioned in under a minute. The same compute backend also powers our Akash provider for container workloads — pay by card on the Akash console.

claim 50% off view specs
< 90s avg provision
1 dc live + expanding
10 tb transfer included
0 egress bandwidth fees
[ 01 / hardware ]
hardware spec

Real silicon.
Not a slide.

Every node is a published datasheet. Same chips, same disks, same network — for everyone.

compute
Intel Xeon
cpu cascadelake at launch
billing per vcpu, per second
shapes compute / memory / storage tuned
memory
ECC
error correction ecc throughout
sizing by sku, no surprise downshifts
consistency uniform fleet-wide
storage
Network NVMe
media all-flash u.2
replication redundant
snapshots 50% of nvme rate
network
50 Gbps / host
per-host uplink 50 gbps
backbone 100 gbe
ipv6 native /64
[ 02 / features ]
features

Built for shipping,
not for slides.

Six things every infra team asks for. We just ship them by default, instead of charging extra.

// 01

Provisioned in 90 seconds

Async provision through the control plane. From API call to root shell in under 90 seconds, every time.

fast
// 02

No egress fees, more DCs coming

DC-1 live in Dallas; additional regions on the roadmap. Bandwidth in and out is free up to your plan's transfer.

network
// 03

Snapshots at half the storage rate

On-demand VM snapshots today, billed at 50% of normal NVMe storage. Cron-style schedules + retention policies ship in v1.x. RWX file shares for shared workloads.

durable
// 04

SSH or API. Pick one.

Full HTTP API with an OpenAPI spec and an official Terraform provider at launch. Or just curl it. Whatever gets you to root.

devx
// 05

Real auth, real audit

MFA, session revocation, scoped API keys, OIDC login, and per-project audit log. Not just a single root token.

secure
// 06

Pricing on the front page

No "contact sales." Same price whether you spin up one VM or fifty. Wallet, statements, and rates all on the API.

honest
[ 03 / compute options ]
compute options

One backend.
Multiple cloud frontends.

The same Metalhost hardware and API powers both centralized (this site, USD billing) and decentralized (Akash, AKT/USDC) cloud frontends. Pick the model that fits your workload.

// 01 · centralized

Virtual machines

Per-second-billed VMs in compute-, memory-, and storage-tuned shapes. Real cores, ECC memory, network NVMe. Standard cloud experience: API + console + CLI. Prepaid wallet — top up by card (Stripe) or USDC (x402).

  • +Per-second billing, no minimums
  • +Prepaid wallet: card or USDC
  • +Snapshots (50% of NVMe rate), audit log, MFA
see pricing →
// 02 · centralized

Bare metal

Whole-machine, single-tenant. Console + rescue mode, your own kernel, full IPMI/iDRAC on a private OOB. Per-second billing scaled to the host SKU. For workloads that need the full machine.

  • +Single-tenant host, no neighbors
  • +Prepaid wallet, per-second billing
  • +Browser console + rescue mode
view hardware →
// 03 · decentralized
new

Containers via Akash

Same hardware, different frontend. We operate a provider on the Akash Network — deploy a container workload by SDL spec and pay by credit card on the Akash console. Bid-based pricing, often cheaper for short-lived workloads.

  • +OCI containers, K8s-style deploys
  • +Pay by card on the Akash console
  • +Bid-based pricing — often cheaper
deploy on akash ↗
One API. Multiple frontends. Metalhost's compute API abstracts hardware provisioning so the same fleet can be sold through centralized billing (this site) and decentralized cloud frontends like Akash. Adding more frontends — or running our software at your own DC — is a roadmap item.
api docs →
[ 04 / pricing ]
pricing

Pay for what you
actually use.

all prices in usd · billed per second · no monthly minimum · cancel any time
tiny
$4.75/ mo
$0.0065 / hr · billed per second
Static sites, bots, learning.
  • + 1 vCPU
  • + 1 GB ECC
  • + 10 GB network NVMe
  • + 10 TB transfer
  • + IPv6 /64 (IPv4 +$2/mo)
spin one up →
medium
$23/ mo
$0.0315 / hr · billed per second
App servers, small databases.
  • + 2 vCPU
  • + 8 GB ECC
  • + 80 GB network NVMe
  • + 10 TB transfer
  • + Snapshots (50% of NVMe rate)
  • + MFA + audit log
spin one up →
large
$46/ mo
$0.0630 / hr · billed per second
Production workloads, CI.
  • + 4 vCPU
  • + 16 GB ECC
  • + 160 GB network NVMe
  • + 10 TB transfer
  • + Snapshots (50% of NVMe rate)
  • + MFA + audit log
spin one up →
bare metal
starting at
$499/ mo
$0.69 / hr · billed per second
Whole machine. Price scales with the host SKU.
  • + Single-tenant host
  • + ECC memory
  • + Local NVMe + network NVMe
  • + 50 Gbps host uplink
  • + Console + rescue mode
  • + Per-second billing
reserve a slot →
Or fully custom. The tiers above are presets. In the dashboard you can dial vCPU, RAM, boot disk, and public IPv4 independently — pick any shape, pay by the second, see a live quote as you tweak. Same per-unit prices for everyone, no contact-sales gating.
configure yours →
[ 05 / waitlist ]
offer first 100 users · 50% off · 3 months

Get on the metal,
at half price.

Public launch june 15, 2026. Drop your email — first hundred founders get half off everything for the first three months. No credit card. No "schedule a call."

waitlist closes when the first 100 fill

// waitlist.signup accepting
$ join_waitlist --offer=beta100
spots claimed 0 / 100