computefare/start match
// full comparison

9 providers, one table.

Every provider ships with a "watch for" caveat because the marketing sites don't.

// methodology
pricing is pulled from providers' own public pricing pages/APIs and rechecked periodically. providers requiring a custom quote are marked as such — we do not guess. last full audit: pending.
showing 9 of 9 providers
provider // digitalocean

DigitalOcean

$—.––
unverified

Developer-first cloud with predictable pricing.

best for
General web hosting, app backends, small GPU inference jobs.
standout
Flat per-hour pricing published on-site; GPU droplets available self-serve.
watch for
GPU capacity is smaller than hyperscalers; H100 availability is regional.
USEUAPACconsumerH100+self-serve
verified: pendingvisit digitalocean
provider // hetzner

Hetzner

$—.––
unverified

Budget dedicated servers with a strong EU footprint.

best for
Cost-sensitive dedicated hosting, long-running training on owned iron.
standout
Price-per-core and price-per-GB routinely the lowest in the market.
watch for
Dedicated (Robot) provisioning is 1–3 days, not instant. GPU stock is limited.
EUUSconsumerself-serve
verified: pendingvisit hetzner
provider // hyperstack

Hyperstack

$—.––
unverified

GPU-focused cloud with EU data centers.

best for
AI training on NVIDIA GPUs at sub-hyperscaler prices.
standout
H100/A100 hourly pricing frequently below the hyperscalers.
watch for
Newer platform; feature depth (networking, storage) is narrower.
EUUSA100H100+self-serve
verified: pendingvisit hyperstack
provider // latitude

Latitude.sh

$—.––
unverified

Self-serve bare metal with a real API.

best for
Teams that want dedicated H100/A100 without a sales call.
standout
API-driven bare metal provisioning in minutes, competitive H100 pricing.
watch for
Smaller region footprint than hyperscalers; capacity can fluctuate.
USEUAPACA100H100+self-serve
verified: pendingvisit latitude.sh
provider // linode

Linode / Akamai

$—.––
unverified

Predictable cloud compute, now backed by Akamai's edge.

best for
Web apps, general backends, some GPU inference.
standout
Flat monthly pricing; Akamai edge locations for latency-sensitive apps.
watch for
Affiliate program status post-Akamai acquisition is unclear — verify.
USEUAPACconsumerself-serve
verified: pendingvisit linode
provider // ovh

OVHcloud

$—.––
unverified

European hyperscaler with dedicated GPU tiers.

best for
Compliance-sensitive EU workloads, dedicated AI training servers.
standout
Broad catalog: bare metal, public cloud, and dedicated H100 nodes.
watch for
UI and docs are dense; some GPU SKUs require a quote.
EUUSAPACconsumerA100H100+self-serve
verified: pendingvisit ovhcloud
provider // scaleway

Scaleway

$—.––
unverified

EU-native cloud with GPU and bare-metal instances.

best for
EU-resident data, mixed GPU + bare metal deployments.
standout
Sovereign EU hosting with H100 instances and bare-metal Elastic Metal.
watch for
Some newer GPU SKUs have waitlist capacity.
EUconsumerA100H100+self-serve
verified: pendingvisit scaleway
provider // vast

Vast.ai

$—.––
unverified

GPU marketplace — rent from independent operators.

best for
Cheapest hourly GPU for training runs where SLAs are optional.
standout
Real-time marketplace; often the lowest $/GPU-hour available.
watch for
Marketplace of independent hosts — reliability and compliance vary per listing.
GlobalconsumerA100H100+self-serve
verified: pendingvisit vast.ai
provider // vultr

Vultr

$—.––
unverified

Global bare-metal and cloud GPU, hourly billing.

best for
Inference, dev/test, bursty workloads across many regions.
standout
Public pricing API; 30+ regions with hourly billing.
watch for
Bare-metal provisioning can be minutes-to-hours depending on region.
USEUAPACconsumerA100H100+self-serve
verified: pendingvisit vultr