Providers

Every vendor in the dataset. Country is where the company is headquartered, which is a reasonable first proxy for billing currency and data residency — but only a proxy: check the region you actually deploy into.

List prices checked

53 providers

ProviderHQServicesModels
AkamaiThe original CDN, which acquired Linode in 2022 to add compute. Enterprise delivery pricing is negotiated, while the Linode-derived cloud products are published and flat-rate.US3
Amazon Web ServicesThe largest cloud provider by revenue, with the broadest service catalogue and the most complex pricing. Rates are rarely the cheapest on any single axis, and egress in particular is priced well above the market.US7
AnthropicThe Claude model family, priced at a premium for frontier reasoning and long-context work, with prompt caching and batch discounts available.US4
BackblazeAn S3-compatible object store priced at roughly a quarter of AWS S3, with free egress up to three times the volume of data stored.US1
bunny.netA Slovenian CDN that competes almost purely on price, with per-region egress rates an order of magnitude below the hyperscalers and no monthly minimum.SI1
CDN77A CDN sold mainly on committed-volume contracts for video delivery, with unmetered options that suit high-bandwidth streaming better than per-gigabyte rates do.GB2
CloudflareAn edge-first network that made zero egress fees its main commercial weapon. Storage and CDN products undercut the hyperscalers specifically on data transfer out.US3
Cockroach LabsA distributed SQL database offering Postgres wire compatibility with multi-region consistency, billed on request units and storage.US1
CohereAn enterprise-focused lab specialising in retrieval, reranking, and multilingual models, sold largely through private deployments.CA2
ContaboA budget host offering unusually large RAM and disk allocations for the money. Performance is variable and the platform is thinner than its competitors'.DE1
CoreWeaveA large specialist GPU cloud aimed at sustained training workloads, priced well under the hyperscalers for the same accelerators.US1
DeepSeekA Chinese lab whose reasoning models are priced an order of magnitude below Western frontier models, with additional off-peak discounts.CN2
DenoDeno Deploy runs JavaScript at the edge with no build step, priced on requests and outbound data with a usable free allowance.US1
DigitalOceanFlat-rate infrastructure aimed at small teams, priced for predictability rather than for the lowest possible unit cost. Bandwidth is pooled across an account's droplets.US4
FastlyA developer-oriented CDN built around instant purging and edge compute. Priced above the discount CDNs, and sold on configurability rather than on cost per gigabyte.US2
Fly.ioRuns containers as lightweight VMs in dozens of regions, billed per second, with a model built around putting the application near the user.US1
GcoreA Luxembourg-based edge provider with a large presence in Europe and Asia, bundling CDN, GPU compute, and object storage under one account.LU1
GoogleGoogle Cloud for infrastructure and the Gemini API for models. Compute pricing is competitive with AWS, and sustained-use discounts apply automatically rather than requiring a commitment.US63
HerokuThe platform that defined the category, now owned by Salesforce. Pricing is high relative to modern alternatives and the free tier ended in 2022.US1
HetznerA German host whose cloud servers are consistently the cheapest per vCPU in the market, with 20 TB of traffic bundled. The trade-off is a small region footprint and no enterprise support tier.DE1
HyperstackA European GPU cloud from NexGen, selling on-demand and reserved accelerators with EU data residency.GB1
IDriveIDrive e2 undercuts most S3-compatible stores on headline storage price and includes egress up to three times stored data.US1
IONOSA German provider bundling cloud servers with hosting and domains, priced for small European businesses and billed per minute.DE1
KeyCDNA Swiss pay-as-you-go CDN with no contracts and per-region pricing. Aimed at small sites that want predictable per-gigabyte billing.CH1
KoyebA serverless platform that deploys containers globally with scale-to-zero, billed per second on provisioned CPU and memory.FR1
LambdaA GPU cloud built for training, selling on-demand and reserved clusters with a hardware business behind it.US1
MicrosoftAzure is the default cloud inside enterprises already licensed for Microsoft software, where existing agreements usually matter more to the final bill than list price does.US6
Mistral AIA French lab shipping both open-weight and commercial models, priced well below the US frontier labs and hosted in the EU.FR3
ModalServerless GPU compute billed per second with scale-to-zero, aimed at bursty inference rather than long training runs.US1
MongoDBAtlas is the managed service for MongoDB, with a permanently free shared cluster and dedicated tiers priced per instance size.US1
NeonServerless Postgres that separates storage from compute and scales to zero, billed on compute-hours rather than on a provisioned instance size.US1
NetlifyA frontend hosting platform with generous build tooling and the most expensive overage bandwidth of the major PaaS options.US2
OpenAIThe most widely adopted model API, with a wide ladder from nano to frontier models and aggressive discounts for cached input tokens.US9
OracleOracle Cloud's headline is an Always Free tier that is genuinely generous by industry standards, plus egress rates well below AWS and Azure.US2
OVHcloudEurope's largest host, operating its own data centres and offering unmetered traffic on most VPS plans — which removes egress from the bill entirely at the low end.FR1
PaperspaceGPU notebooks and machines now owned by DigitalOcean, aimed at individual researchers and small teams rather than at cluster-scale training.US1
PlanetScaleManaged MySQL built on Vitess, sold on branching workflows and horizontal scale. Pricing starts higher than most managed databases and targets production teams.US1
RailwayUsage-based hosting that bills per second of actual CPU and memory consumption, which suits side projects and irregular workloads.US1
RenderA Heroku-style platform with predictable per-instance pricing, managed Postgres, and free static hosting.US1
ReplicateRuns open models behind a hosted API billed per second of GPU time, trading unit price for not having to manage anything.US1
RunPodA GPU marketplace splitting capacity between vetted datacentres and a cheaper community tier, billed per second with no commitment.US1
ScalewayA French cloud with EU data residency as its main pitch, covering compute, object storage, and managed databases at rates between Hetzner and the hyperscalers.FR3
StorjDistributed object storage that spreads erasure-coded shards across independent operators, priced flat worldwide with no per-region tiers.US1
SupabaseA Postgres platform bundling auth, storage, and realtime subscriptions, priced per project rather than per resource.SG1
Together AIServes open-weight models as an API and rents dedicated GPU clusters, positioned between a model provider and a GPU host.US1
TursoSQLite replicated to the edge, priced on row reads and writes, aimed at read-heavy applications that want the database close to the user.US1
UpCloudA Finnish host selling on storage I/O performance, with per-instance transfer allowances and a simpler product surface than the hyperscalers.FI1
UpstashServerless Redis and Kafka billed per request rather than per hour, which suits bursty and edge workloads that would otherwise pay for idle instances.US1
Vast.aiA peer-to-peer GPU marketplace where independent hosts bid capacity down. The cheapest rates anywhere, with reliability that varies by host rather than by SLA.US1
VercelThe managed platform for Next.js, priced per seat plus usage. Convenience is the product; bandwidth costs several times what a plain CDN charges.US2
VultrA price-competitive VPS host with an unusually wide region list, including locations the larger clouds do not serve. Bandwidth is allocated per instance.US1
WasabiFlat-rate hot object storage with no egress or API request charges, subject to a policy that egress should not exceed stored volume and a 90-day minimum retention.US1
xAIThe Grok model family, sold with large context windows and pricing positioned between the mid-tier and frontier offerings of its competitors.US2