Cloud provider accounts, listed plainly and priced in US dollars.

AWS runs hundreds of services, but most accounts rely on a familiar core: EC2 virtual machines, S3 object storage, RDS managed databases, Lambda functions, VPC networking, IAM and CloudFront for content delivery. Compute capacity is governed by service quotas. EC2 On-Demand quotas are counted in vCPUs per region and per instance family, so a quota describes how many cores can run at once in one region, not how many instances exist. Amazon Bedrock exposes foundation models through an API with per model quotas for requests and tokens per minute. Kiro is the agentic development environment from AWS.

A startup can build an API on Lambda, DynamoDB and API Gateway and pay only for requests. A data team can run a nightly Spark job on EMR against S3 and shut it down. A research group can train on GPU instances for a week and archive results to Glacier. A QA team can copy a production VPC into a separate account and test a migration without touching live systems. Each of those needs a different quota, which is why the tiers on this listing are separated by vCPU count rather than by instance count.

• vCPU tier labels describe the listed On-Demand quota as labeled; AWS sets and adjusts actual limits per region and per instance family.
• Credits Account labels describe a listed promotional balance; AWS applies promotional credit only to eligible services and until it expires.
• Bedrock Enabled and Kiro Enabled labels describe listed access to those services; model availability differs by region and is set by AWS.
• AWS Free Tier terms are defined by AWS and change over time; check the current offer on the AWS Free Tier page before relying on it.

Google Cloud organises resources into projects, each attached to a billing account that pays for its usage. Compute Engine provides virtual machines, Google Kubernetes Engine runs containers, and Cloud Run runs containers without managing servers. Data is where Google Cloud is strongest: BigQuery runs SQL over very large datasets with separate pricing for storage and queries, Cloud Storage holds objects in regional or multi-regional buckets, and Pub/Sub and Dataflow handle streaming. Vertex AI brings model training, tuning and serving together, including Gemini and third party models through Model Garden. Quotas are enforced per project and per region.

A product analytics team can stream events through Pub/Sub into BigQuery and query them the same minute. A machine learning engineer can fine tune a model in Vertex AI, deploy an endpoint and call it from Cloud Run. A startup can host its API on Cloud Run, its database on Cloud SQL and its files in Cloud Storage while paying by the second. A researcher can rent TPU or GPU time for a single experiment and release it afterwards. All of these are billed against the project’s billing account.

• Prepaid labels such as $300 + $10 Prepaid describe a listed trial balance and top-up only; Google sets trial terms and they are subject to change.
• Credits Account labels describe a listed credit balance as labeled; Google applies credit to eligible SKUs on the linked billing account.
• Buyers searching for a $400 credit option will not find that label here; the closest listed variants are in the Prepaid and Credits groups, and support can confirm what is currently available.
• Vertex AI model access, regional availability and rate limits are controlled by Google, not by the account seller.

Azure is organised around subscriptions, which hold resource groups, which hold resources. Identity comes from Microsoft Entra ID, so the same directory that signs users into Microsoft 365 can control access to cloud resources. Core services include Virtual Machines, Azure Kubernetes Service, App Service for web apps, Azure Functions, Azure SQL Database, Cosmos DB and Blob Storage. Azure AI Foundry is the portal for building with hosted models, including the Azure OpenAI Service. Compute is limited by regional vCPU quotas per VM family, and support is sold in tiered plans that set response times.

An enterprise IT team can lift Windows Server workloads into Azure VMs and keep their existing licences under Hybrid Benefit. A .NET team can deploy from GitHub or Azure DevOps to App Service with staging slots. A startup can build a chat assistant on Azure OpenAI Service with data kept in Azure AI Search. A test team can spin up a full environment with Bicep or Terraform, run acceptance tests, then delete the resource group in one command. A university can give students separate subscriptions under a single Entra tenant.

• The $200 Credit labels describe a listed credit as labeled; Microsoft sets the amount, validity period and eligible services for its own free credit.
• The Azure $1K Tier 1 Support Claude label is listed exactly as written; it does not add any capability beyond what the label states.
• Regional vCPU quotas start low on new subscriptions and increases are requested through the Azure portal.
• AI model availability in Azure AI Foundry varies by region and is controlled by Microsoft.

A Droplet is a Linux virtual machine billed hourly up to a fixed monthly cap. Plans range from Basic shared CPU through General Purpose, CPU-Optimized and Memory-Optimized to GPU Droplets. Around the Droplet sit Volumes for block storage, Spaces for S3-compatible object storage, Managed Databases for PostgreSQL, MySQL, MongoDB and Kafka, DigitalOcean Kubernetes, Load Balancers, VPC networking and Cloud Firewalls. App Platform deploys from a Git repository without managing servers. Data centers are in New York, San Francisco, Toronto, Amsterdam, London, Frankfurt, Bangalore, Singapore and Sydney. Every account has a Droplet limit that DigitalOcean raises on request.

A freelancer can host a dozen client WordPress sites on separate Droplets and snapshot each one before updates. A SaaS startup can give each tenant an isolated Droplet and manage the fleet through the API and Terraform. A development team can run staging on a small Droplet and production behind a Load Balancer. A student can learn Linux administration on the cheapest plan. The Open 25 Port options are aimed at applications that send legitimate transactional and notification email in line with the provider’s acceptable use policy.

• The 3 Droplet and 10 Droplet labels describe the listed account limit as labeled; DigitalOcean sets and adjusts the actual limit.
• DigitalOcean’s SMTP policy explains that outbound port 25 is restricted on new accounts by default, so check the listed label before ordering.
• Email sent from any Droplet must follow DigitalOcean’s terms and anti-spam law; opt-in recipients and a working unsubscribe are expected.
• Managed Databases and Spaces are billed separately from Droplets and count against their own limits, so plan the full stack, not just the Droplet count.

Linode compute instances come in Shared CPU, Dedicated CPU, High Memory, Premium and GPU plans, each with a fixed monthly price and bundled outbound transfer. Block Storage volumes attach to instances, Object Storage is S3-compatible, and the Linode Kubernetes Engine runs managed clusters. NodeBalancers distribute traffic, VLANs and VPCs provide private networking, and Cloud Firewalls filter at the edge of the instance. Since the acquisition, Akamai has added core and distributed regions in more countries and connected the platform to its edge network for delivery and security. Management is through the Cloud Manager console, the Linode CLI and Terraform.

A web agency can host client sites on Shared CPU plans and move the busiest one to Dedicated CPU without rebuilding. A game studio can place match servers in regions near its players. A startup can run its application on LKE and put static assets behind Akamai delivery. A developer can use the listed $100 credit to compare Linode against DigitalOcean or Hetzner on the same workload. The Open 25 Port option is intended for legitimate transactional and notification email that follows Akamai’s acceptable use policy.

• The Free Trial $100 Credit label describes a listed promotional credit as labeled; Akamai sets the terms and expiry and they are subject to change without notice.
• Linode’s email sending guide describes its default outbound mail restrictions for new accounts and the reverse DNS and SPF records it expects before lifting them.
• Plans include a monthly outbound transfer allowance pooled across the account, with overage billed per gigabyte beyond the pool.
• Backups are an optional paid add-on per instance; snapshots and images are managed separately and can be used to clone a server into another region when needed.

OVHcloud sells four infrastructure families. Public Cloud is OpenStack based and offers instances, block and object storage, managed Kubernetes and databases, billed hourly or monthly under per-project quotas for vCPUs, RAM and instances. VPS is a fixed-price virtual server line. Bare Metal servers are physical machines with no hypervisor, from entry level to GPU and storage configurations. Hosted Private Cloud runs VMware on dedicated hardware. All lines connect through vRack, OVHcloud’s private network, and include anti-DDoS protection. Data centers are concentrated in France, Germany, Poland and the United Kingdom, with sites in Canada, the United States, Singapore, Australia and India.

A European SaaS company can keep customer data in a French or German region for GDPR reasons. A media company can transcode on dedicated servers and store output in object storage. A development team can use the 34 vCPUs quota for a Kubernetes cluster with several node sizes. A hosting reseller can rent VPS plans and manage them through the API. A university lab can rent a GPU dedicated server for a semester. The Open 25 Port option is intended for legitimate transactional and notification email that respects OVHcloud’s acceptable use policy.

• The 34 vCPUs label describes a listed Public Cloud quota as labeled; OVHcloud adjusts project quotas on request through its control panel, and quotas apply per region.
• OVHcloud is subject to European law, and some of its services hold the SecNumCloud qualification used by the French public sector.
• Dedicated servers are ordered per month, some ranges carry setup fees, and delivery times vary by model and location.
• Outbound mail from OVHcloud infrastructure must comply with its terms and with anti-spam law in the recipient’s country, including consent and unsubscribe requirements.

OCI puts everything inside a tenancy, split into compartments for access control. Compute uses flexible shapes with OCPUs and memory chosen independently, on AMD, Intel and Ampere Arm processors, plus bare metal and GPU shapes. Database services are the main draw: Autonomous Database, Exadata Database Service, Base Database Service and MySQL HeatWave. Oracle Kubernetes Engine, block, file and object storage and virtual cloud networks cover the rest. Oracle Cloud Free Tier has two parts: a time-limited trial credit, and Always Free resources such as Arm compute, Autonomous Databases and storage that outlast the trial. Oracle sets and changes these terms.

A team running Oracle Database on-premises can move it to Autonomous Database and stop patching. A developer can run a small application on an Always Free Arm instance for the long term. A startup can test MySQL HeatWave for analytics without a separate warehouse. A research group can rent GPU shapes for one training run. A consultancy can keep one tenancy per client and separate teams with compartments. Upgrading to Pay As You Go removes the trial’s service limits and lets a tenancy grow beyond Always Free quotas.

• The Free Trial $300 label describes a listed trial balance as labeled; Oracle sets its trial amount, duration and eligible services, and they are subject to change.
• The Upgraded Pay As You Go label describes a listed billing state only; usage beyond free allocations is charged by Oracle at its published rates.
• Service limits apply per tenancy and per region; increases are requested through the OCI console and reviewed by Oracle.
• Always Free resources are subject to capacity in the home region chosen at signup, and the home region cannot be changed later.

Hetzner Cloud servers come in shared vCPU plans on Intel, AMD and Ampere Arm processors and in dedicated vCPU plans, all billed hourly with a monthly cap. Around them are Volumes, Load Balancers, Floating IPs, private Networks, Firewalls, Snapshots, Backups and Object Storage. The console groups resources into projects, and every project has a server limit that Hetzner raises on request. Hetzner also sells dedicated root servers, a Server Auction of discounted hardware, and Storage Box backup space. Locations are Falkenstein, Nuremberg, Helsinki, Ashburn, Hillsboro and Singapore, with larger traffic allowances in Europe than elsewhere.

A developer can run a Kubernetes cluster on five small servers. A startup can host its API in Falkenstein, its database in Nuremberg and connect them over a private Network. A team can build CI runners that are created and destroyed through the API or the hcloud CLI. A self-hoster can run Nextcloud, a mail client and a password manager on dedicated vCPU plans. The Open 25 Port option is intended for legitimate transactional and notification email that complies with Hetzner’s acceptable use policy.

• The 5 Server and 10 Server labels describe the listed project limit as labeled; Hetzner controls the actual limit and increases it through a request in the console.
• Hetzner’s Cloud FAQ explains that outbound mail ports are blocked on new accounts by default, so check the listed label before ordering.
• Buyers searching for a Hetzner free trial option will not find that label here; the closest listed options are the server limit accounts.
• Arm based plans are cheaper than x86 plans at the same size but require Arm compatible software images such as Ubuntu or Debian Arm builds.

New SES identities start in a sandbox where you can only send to verified addresses and with a small daily quota. Production access is requested through the AWS console and, once granted, AWS assigns a daily sending quota and a maximum send rate that it can raise over time based on sending behaviour. Mail can be submitted through the SES SMTP interface using generated SMTP credentials, or through the SES API and the AWS SDKs. Configuration sets, dedicated IP options, event publishing to SNS or CloudWatch, and DKIM signing for verified domains are all part of the service.

A typical setup is a web application that sends signup confirmations, password resets, invoices and order updates through SES using the SMTP interface from a framework mailer such as Django, Laravel or Nodemailer. Startups often route all product notification email through SES and keep marketing lists on a separate opt-in platform. Test environments use the sandbox with verified recipients before any production sending. In every case, sending must comply with AWS’s SES terms and anti-spam law, including CAN-SPAM, GDPR and the equivalent rules in the recipient’s jurisdiction.

• The SMTP 50K Limits and SMTP 100K Limits labels describe the listed sending limit only; AWS sets and adjusts the actual quota on every SES account.
• SES requires verified sender identities: a domain with DKIM records published in DNS, or an individual email address confirmed by link.
• AWS monitors bounce and complaint rates; sustained high rates can lead to a sending pause or review under SES policy.
• Bulk or marketing email is only acceptable to recipients who opted in, with a working unsubscribe mechanism in every message.

UpCloud’s compute product is the Cloud Server, available in preconfigured plans and in custom configurations where CPU and memory are chosen separately. Storage uses MaxIOPS, UpCloud’s own block storage layer built for high input/output performance, alongside slower archive-class storage. Data centres are located in Europe (including Helsinki, Frankfurt, Amsterdam and London), the United States (including Chicago and New York), Asia (Singapore) and Australia (Sydney). Private SDN networking, floating IPs, managed databases, object storage and a managed Kubernetes service are also offered, with a REST API and a Terraform provider for automation.

Common workloads include database-backed web applications where MaxIOPS storage keeps PostgreSQL or MySQL latency low, CI runners that create and destroy servers through the API, and distributed test nodes across Europe, the US, Asia and Australia. Startups use the credit to prototype a full stack, including a managed database and object storage, before committing to a monthly budget. Agencies host client staging sites in a region near the client and copy the setup to production with Terraform. Because billing is hourly, a team can build a staging copy of production on a Monday, test against it all week, and delete it on Friday without paying for a full month of servers.

• Billing is hourly with a monthly cap per resource; the $500 Credit variant is listed as labeled and is drawn down by usage.
• MaxIOPS volumes can be resized and snapshotted, and scheduled backups are configured per server.
• Server plans can be changed after deployment, so start small and scale once the workload profile is clear.
• Credit terms and any promotional conditions are set by UpCloud and are subject to change.

IBM Cloud Kubernetes Service and Red Hat OpenShift on IBM Cloud are the two managed container platforms, with worker pools sized by vCPU and memory and placed across multi-zone regions such as Dallas, Washington DC, London, Frankfurt, Tokyo and Sydney. Virtual Private Cloud (VPC) provides virtual server instances, block storage and load balancers. The watsonx family covers watsonx.ai for model building and prompting, watsonx.data for lakehouse storage and watsonx.governance for AI lifecycle controls. Cloud Object Storage, Db2, Cloudant, Code Engine for serverless containers, and IAM with resource groups and access policies complete the platform, and many services offer a Lite plan with a free usage allowance.

Typical scenarios include a team deploying a Spring Boot or Node.js microservice to a small Kubernetes cluster to compare cluster management with EKS or GKE, a data science group running prompt experiments on foundation models in watsonx.ai, and a developer testing Code Engine for a scheduled batch job. Enterprises with existing IBM software licences use the trial to validate a migration path for Db2 or MQ workloads. Trial scope, limits and duration are set by IBM and are subject to change.

• The IBM Cloud Free Trial variant is listed as labeled; IBM defines what the trial includes and when it converts to a paid account.
• IBM Cloud account types (Lite, Trial, Pay-As-You-Go and Subscription) determine which services and quotas are available.
• Kubernetes and OpenShift clusters are billed per worker node hour; delete unused clusters to avoid charges.
• Set up IAM resource groups and access groups before inviting collaborators so permissions stay scoped.

Elastic Compute Service (ECS) is the core compute product, with instance families for general, compute-optimised, memory-optimised and GPU workloads, pay-as-you-go or subscription billing, and images from the Alibaba Cloud Marketplace. Object Storage Service (OSS) provides S3-compatible object storage with lifecycle rules and CDN integration. ApsaraDB covers managed RDS for MySQL, PostgreSQL and SQL Server, plus Redis and MongoDB. Container Service for Kubernetes (ACK), Function Compute, Server Load Balancer, VPC and Alibaba Cloud CDN complete the stack. The region map spans mainland China, Hong Kong, Singapore, Jakarta, Kuala Lumpur, Tokyo, Seoul, Sydney, Mumbai, Dubai, Frankfurt, London, Silicon Valley and Virginia.

A cross-border e-commerce startup might run its storefront on ECS in Singapore with OSS for product images and an ApsaraDB RDS instance, then add a Hong Kong region later for lower latency to southern China. Game studios test matchmaking servers in Tokyo and Seoul. Data teams load CSV exports into MaxCompute or AnalyticDB for reporting. Development teams use the trial to script ECS provisioning with the Alibaba Cloud CLI or Terraform before rolling out to production. Trial contents are decided by Alibaba Cloud and change over time.

• The Alibaba Cloud Free Trial variant is listed as labeled; eligible products and duration are set by Alibaba Cloud.
• Hosting a public website in a mainland China region requires an ICP filing; regions outside mainland China do not.
• The international console at alibabacloud.com and the China site at aliyun.com are separate account systems.
• Pay-as-you-go ECS instances are billed for as long as they exist; stop or release instances you no longer need.

A Workspace tenant is tied to a verified domain and administered from admin.google.com, where an admin creates users, groups and organisational units, assigns licences, and sets security policies such as 2-Step Verification enforcement and third-party app access. Gmail on a custom domain includes spam filtering, routing rules and SPF, DKIM and DMARC configuration. Drive provides per-user storage and shared drives; Docs, Sheets and Slides support real-time co-editing; Meet handles video calls; Calendar manages shared rooms and resources. Plans include Business Starter, Standard and Plus, and Enterprise tiers with Vault and advanced endpoint management. Google for Nonprofits is a separate programme through which eligible organisations can obtain Workspace at no or reduced cost.

A five-person startup buys a domain, points its MX records at Google, and has firstname@company.com inboxes, shared calendars and a Drive folder structure the same day. An agency creates a shared drive per client and uses Groups for role addresses such as support@ and billing@. Developers use Workspace accounts to reach Google Cloud projects with organisation-level IAM and to configure OAuth consent screens for internal apps. Charities that qualify for Google for Nonprofits use that programme’s Workspace edition for staff and volunteers; eligibility is determined by Google, not by this marketplace.

• Gsuite Workspace Inbox Supported and Gsuite Workspace Non Profit 300 are listed exactly as labeled; the product page states what each includes.
• The Non Profit 300 label is the listed variant name only; Google decides eligibility for Google for Nonprofits and can review status at any time.
• Domain ownership verification through a DNS TXT record is required before Gmail is active on the domain.
• Configure super admin recovery options and 2-Step Verification before adding users to the tenant.

Kamatera servers are built from individual settings: you pick the number of vCPUs, the CPU type (Availability, General Purpose, Dedicated or Burstable), RAM in gigabyte steps, SSD storage size, and the operating system or application image, and the price updates per component. Servers can be billed hourly or monthly with a traffic allowance, and resources can be raised or lowered after deployment. Data centers include New York, Toronto, London, Amsterdam, Frankfurt, Tel Aviv, Hong Kong and Singapore. Private VLANs, load balancers, firewall management, daily backups, block storage and a REST API for provisioning are available from the same console.

A common case is a memory-heavy application such as Elasticsearch or a Java service where a developer wants 2 vCPUs with 16 GB of RAM rather than the pairing a fixed plan would impose. Agencies use the Free Trial variant to stage a WordPress or Magento build and measure response time from a nearby data center. Teams that self-host Postfix or a ticketing system that sends notifications use the Port 25 Open variant to deliver opt-in transactional messages with SPF, DKIM and reverse DNS configured.

• Kamatera Free Trial is listed as labeled; trial scope and duration are decided by Kamatera and subject to change.
• Kamatera Port 25 Open is listed with outbound port 25 available as labeled; sending must follow Kamatera’s acceptable use policy and anti-spam law.
• Hourly billing suits short tests, while monthly billing usually bundles a traffic allowance for steady workloads.
• Managed services and extended support are optional paid add-ons on top of the server price.

Vultr Cloud Compute comes in Regular (shared vCPU), High Frequency, High Performance and Optimized Cloud Compute families with dedicated vCPUs, all deployable from the customer portal or the API. Bare Metal gives single-tenant physical servers on the same hourly model. Cloud GPU offers NVIDIA accelerators, including fractional GPU slices, for inference and training. Vultr Kubernetes Engine, Managed Databases for MySQL, PostgreSQL and Kafka, Block Storage, Object Storage, Load Balancers, VPC networking and DDoS protection are available in most regions. Locations span North America, South America, Europe, Asia, Australia and Africa.

A developer with a Vultr 250$ Credit variant can run a High Frequency instance for a Node.js API with a managed PostgreSQL database and a load balancer while the product finds its first users. Machine learning students use a fractional Cloud GPU slice for fine-tuning runs that are too slow on a laptop. Teams already hosting on DigitalOcean or Linode add a Vultr region in Johannesburg, Sao Paulo or Mumbai to serve users there. Test pipelines create and destroy instances by API and pay only for the hours used.

• Six credit variants are listed, 200$, 205$, 250$, 255$, 300$ and 305$ Credit, each as labeled; searches for a Vultr free trial account are best matched by the smallest credit option.
• Vultr credit is consumed by hourly usage across compute, storage, bandwidth overage and add-ons.
• Credit terms, expiry and promotional conditions are set by Vultr and can change.
• Snapshots and automatic backups carry their own charges on top of instance pricing.

BinaryLane cloud servers are KVM virtual machines with SSD storage, configurable vCPU, memory and disk, and a choice of Linux distributions or Windows Server images. Servers run in Sydney, Melbourne, Brisbane and Perth, so a workload can be placed in the same city as its users or spread across states for redundancy. Features include IPv4 and IPv6 addressing, private VPC networking, firewall rules, scheduled backups, snapshots and a REST API with a published specification. Pricing is quoted in Australian dollars, which simplifies local invoicing for Australian companies and keeps exchange rate movement out of the monthly bill.

A Sydney web agency hosts client sites in Sydney and keeps a warm standby in Melbourne. A Perth engineering firm runs an internal ERP on a server in the same city as its staff. Developers use the Free Trial variant to measure disk and network throughput from Australian ISPs before migrating from a US region. Teams using the Port 25 Open variant run Postfix or a self-hosted helpdesk that sends ticket notifications, with SPF, DKIM, DMARC and reverse DNS configured so opt-in mail is accepted by receiving servers.

• Binarylane Cloud Free Trial is listed as labeled; trial conditions are set by BinaryLane and subject to change.
• Binarylane Cloud Port 25 Open is listed with outbound port 25 available as labeled; use is limited to compliant, opt-in email under BinaryLane’s acceptable use policy and Australia’s Spam Act.
• Four Australian data centres help meet the data residency expectations of local clients and government tenders.
• BinaryLane bills in Australian dollars on its side; the listing prices on this marketplace are in US dollars.