Every price and every signal we publish originates from a marketplace partner integration. Each partner receives identical treatment, identical coverage, and identical standards. No exceptions.
Akash is an open-source, decentralized compute marketplace built on the Cosmos SDK. Providers bid to fulfil containerized (Kubernetes) deployments, anyone with compatible hardware can join as a provider, and settlement runs on-chain. Each deployment is described in a YAML file (SDL) that sets the container image, resources and the highest price the customer will pay. Independent providers respond with bids, and accepting one creates a lease that fixes the price, with payment made per block for as long as the deployment runs. AKT, the network's native token, is used for staking, governance and transaction fees.
AIMC reads Akash GPU prices daily from the public Akash Console pricing API, which reports pricing across the provider network by GPU model.
RunPod is a GPU cloud for machine-learning workloads, offering on-demand GPU instances in Community Cloud and Secure Cloud tiers alongside serverless GPU endpoints. Its catalog runs from consumer RTX cards to data-center H100 and H200 systems. It was started by Zhen Lu and Pardeep Singh, who first shared the platform with developers in early 2022. Pods provide on-demand GPUs, Serverless offers API-based GPU endpoints, and Clusters provide multi-node GPUs for distributed workloads. Compute is billed per second from prepaid credits with no data-transfer fees, and RunPod states ISO/IEC 27001 certification and a completed SOC 2 Type II report.
AIMC reads RunPod's published per-GPU-hour rates daily through its GraphQL API.
TensorDock is a marketplace for GPU and CPU cloud compute, used for AI, deep learning and rendering workloads, with capacity supplied by independent hosting providers. It has been part of Voltage Park, a GPU cloud provider, since March 2025 and continues to run as its own marketplace. Collection is paused: TensorDock's public marketplace has returned no listings since mid-June 2026. Prices observed before the pause remain in the historical record.
No current public listings to observe. AIMC re-checks TensorDock periodically and resumes collection when public listings return.
Verda (formerly DataCrunch) is a European GPU cloud operating from European data centers. It offers on-demand NVIDIA GPU instances and bare-metal options for AI training and inference, including B200 and H200 hardware. Founded in 2020 by Ruben Bryon and headquartered in Helsinki, it runs data centers in Finland and Iceland, with its Helsinki sites powered by hydroelectric, wind and other renewable energy. Alongside GPU instances, it offers instant clusters, serverless containers and inference services.
AIMC reads Verda's published on-demand instance prices daily through its instance-types API.
SaladCloud is a distributed GPU network that runs containerized workloads on idle consumer GPUs contributed by individual PC owners, with a managed orchestration layer handling node selection and failover. Its catalog is mostly consumer GeForce hardware, suited to inference, rendering and batch jobs that can run in parallel. Contributors, whom Salad calls Chefs, run its desktop app while their machines are idle and earn rewards for the capacity they share. Workloads run as containers through the Salad Container Engine, which moves a job to another node of the same GPU class if one drops out. Billing is per second with no contracts or minimums, and reserved data-center hardware is offered separately as Salad Dedicated.
AIMC reads SaladCloud's published per-GPU-hour rates daily through its API.
io.net is a decentralized physical infrastructure network (DePIN) that pools GPU capacity from data centers, mining operations and individual providers, and can group GPUs across locations into clusters. Its catalog spans NVIDIA data-center hardware from A100 through H200 and B200. The network launched in June 2023 and runs on the Solana blockchain. Hardware owners supply GPUs through IO Worker, customers deploy virtual machines, containers and clusters through IO Cloud, and the IO token is used to pay for compute and reward suppliers. Typical workloads include batch inference, model serving and parallel training.
AIMC reads io.net's listed GPU prices daily through its API.
Cirrascale is a San Diego-based bare-metal GPU cloud offering dedicated multi-GPU servers with NVIDIA (H100, H200, B200) and AMD Instinct accelerators, managed networking and custom configurations for training and inference. The company began as Cirrascale Corporation, a builder of blade and rackmount compute systems, launched its multi-GPU cloud service in 2015, and became Cirrascale Cloud Services in 2017 after selling its hardware business. It describes itself as a private AI cloud and managed services provider, also offers Qualcomm Cloud AI and Tenstorrent accelerators, and bills at a flat monthly rate with no data ingress or egress fees.
Collection is paused: since 5 November 2026, Cirrascale's pricing page shows "as low as" rates without a stated commitment term or GPU count, so AIMC does not publish them until Cirrascale confirms the basis. Monthly committed-term prices observed before the pause remain in the historical record.
Massed Compute is a GPU cloud, founded in 2021, that owns and operates its own infrastructure in Tier III data centers. It rents NVIDIA hardware from A30 and RTX A6000 through H100, H200 and RTX PRO 6000 Blackwell by the hour, with no long-term contract. The company says it buys its own hardware and runs its own network, with no sub-leased server capacity, and its own engineers manage the servers. It offers virtualized and bare-metal GPU instances, GPU clusters and an API, on a pay-as-you-go or contract basis, and states SOC 2 Type II, HIPAA and GDPR compliance.
AIMC reads Massed Compute's published hourly rates daily through its API.
Transparent GPU pricing is critical for enterprise leaders and AI developers. We're proud to be a part of this marketplace ecosystem and congratulate AIMC on bringing data-driven clarity and visibility.
”Vast.ai is a peer-to-peer GPU marketplace where independent hosts list their hardware and set their own prices. It offers on-demand and interruptible instances across most major GPU architectures, from consumer cards to multi-GPU data-center servers. On-demand instances run at the host's fixed price, interruptible instances are allocated by bidding and are paused when outbid, and reserved instances give prepaid discounts on on-demand rates. Co-founded by Jake Cannell in 2016 and headquartered in Los Angeles, with an engineering office in San Francisco, Vast.ai also offers clusters, serverless inference and a Secure Cloud tier run in certified partner data centers.
Peer-to-peer model: independent hosts set their own prices, producing wide price dispersion across thousands of listings. AIMC collects Vast.ai in two passes each day — available listings and currently rented listings — so the gap between asked prices and the prices of machines that actually rented is observable.
Hyperstack, operated by NexGen Cloud, is a GPU-as-a-Service cloud with data centers in North America and Europe. It offers on-demand, reserved and spot NVIDIA GPU VMs from RTX A4000 through H100 SXM, H200 and B200. Launched in 2023 by London-based NexGen Cloud, which was founded in 2020, it runs regions in Norway, Canada and the United States, and describes its Norwegian and Canadian regions as sustainably powered. Usage is billed per minute, spot VMs run on surplus capacity at a fixed discount and can be reclaimed without notice, and the platform also offers on-demand Kubernetes, object storage and its AI Studio service.
AIMC reads Hyperstack's hourly rates daily through its Pricebook API.
Clore.ai is a peer-to-peer GPU marketplace where server owners list their hardware and set their own prices. It offers on-demand and spot rentals across consumer, prosumer and data-center GPUs. Renters pay in cryptocurrency, with deposits accepted in CLORE, BTC, USDT and USDC. On-demand rentals run at a fixed price and cannot be interrupted, while spot rentals are won by bidding, carry lower fees and end if another renter outbids them. Holding CLORE, the platform's native token, reduces marketplace fees, and typical workloads include AI training, inference and 3D rendering.
AIMC reads Clore.ai's node-level listings daily through its documented public API, which returns GPU model, specifications, price and rental status for each server.
Hugging Face is a New York-headquartered open-source AI platform, founded in 2016. Its Inference Endpoints product offers dedicated GPU deployments hosted on AWS and Google Cloud in US and EU regions, from single T4 instances to 8x H200 systems, at fixed published prices. The company is best known for the Hugging Face Hub, where developers share models and datasets, and for its Transformers library. Inference Endpoints deploys a Hub model in a prebuilt container using engines such as vLLM, Text Generation Inference and llama.cpp, with autoscaling and scale-to-zero, and prices are listed per hour and billed per minute.
AIMC reads Inference Endpoints prices daily from Hugging Face's public provider API. These are managed-service prices that include the platform layer on top of the underlying cloud hardware, so they are not directly comparable to bare GPU rental rates.
Trooper.AI is a German GPU server rental platform, founded in 2022, running refurbished enterprise hardware in data centers in Germany and the Netherlands. Its catalog lists individual bare-metal servers, from RTX 4090 and RTX 5090 cards to A100 SXM and RTX PRO 6000 Blackwell, billed in hourly, weekly or monthly intervals and priced in euros. Operated by Media Trooper GmbH of Idstein, Germany, a member of the Cloud Services Made in Germany initiative, it describes its approach as up-cycling high-end used hardware. Each server comes with full root access and one-click templates for tools such as vLLM, ComfyUI and Jupyter, and the company offers EU hosting with a GDPR data-processing agreement.
AIMC reads Trooper.AI's node-level listings daily from its public API and converts EUR prices to USD at the ECB reference rate.
OpenRelay (formerly VectorLay) is a GPU compute marketplace offering on-demand NVIDIA and AMD data-center and consumer hardware, including H100 SXM, B200, B300, MI355X, A100 PCIe 80GB, RTX PRO 6000 Blackwell, RTX 5090, RTX 4090 and RTX 3090, each at a single published per-GPU-hour rate. Based in San Francisco and founded by Jaden Wang and Prashant Patel, it describes itself as distributed, hardware-agnostic AI inference: GPU owners contribute capacity to its network, and OpenRelay schedules customer workloads onto it with automatic failover. GPUs are rented as Ubuntu virtual machines with root access and direct GPU passthrough, billed by the second with no minimum commitment, and storage, networking and data egress are included.
AIMC reads OpenRelay's published rates daily through its pricing API.
Oblivus is a GPU cloud provider with multi-tenant data centers in Montreal and Oslo and further sites in the United States. It offers on-demand and reserved NVIDIA GPU instances (H100 SXM5, H100 and A100 in PCIe and NVLink configurations, L40, RTX 4090, RTX A6000), with vCPU, RAM and storage included in the published hourly rate and usage billed by the minute. The company was established as Oblivus Limited in March 2022 and has operated as Oblivus, Inc. in the United States since May 2024. Instances are virtual servers with dedicated GPU passthrough, on-demand use carries no contract, reservations are discounted according to their length and size, and data ingress and egress are free.
AIMC reads Oblivus's on-demand and reserved per-GPU-hour rates daily from its public pricing page. Because vCPU, RAM and storage are bundled into the rate, these are not bare-GPU prices. Storage, bandwidth and IP pricing are out of scope.
Fal.ai (Features & Labels, Inc.) is a San Francisco-based AI cloud for generative-media inference, founded in 2021, offering on-demand NVIDIA GPU compute billed by the hour (H100 SXM, H200 SXM, B200 SXM, B300 SXM, RTX PRO 6000 Blackwell). Founded by Burkay Gur and Gorkem Yurtseven, the company is best known for its model APIs, which give developers access to generative-media models, and for serverless deployment of customers' own models. Its Compute product, the part AIMC observes, provides dedicated GPU instances with full SSH access at fixed hourly rates for sustained work such as training and fine-tuning, with InfiniBand networking between eight-GPU instances.
AIMC reads Fal.ai's on-demand per-GPU-hour list prices daily from its public pricing page. Its per-token generative-media model API is a separate product and is out of scope.
Exoscale (Akenes SA) is a Swiss cloud provider founded in 2011 and based in Lausanne, offering GPU compute in data centers in Switzerland, Austria, Germany and Croatia, billed by the second with published per-zone pricing. It is a member of A1 Digital, part of the A1 Telekom Austria Group, and its wider cloud also runs zones in Bulgaria and Spain. GPU servers offer one to eight NVIDIA cards per instance with direct passthrough, alongside managed Kubernetes, object storage and managed databases, and the company lists ISO 27001, BSI C5 and GDPR among its compliance frameworks.
AIMC reads Exoscale's per-zone GPU prices daily from its public pricing API. Rates are per instance, so figures include the bundled vCPU and RAM rather than a bare-GPU rate. Its B300 is quote-only and therefore not shown.
QuantaCloud (Quanta Cloud LLC) offers on-demand NVIDIA GPU instances through a self-serve console, billed by the second, alongside reserved and dedicated capacity for production workloads. It describes itself as a GPU cloud for training and inference. Reserved capacity is available as dedicated nodes and multi-node deployments, with InfiniBand networking offered for reserved clusters, and its lineup also includes dedicated GPU servers and GPU virtual private servers.
AIMC reads QuantaCloud's full on-demand catalog daily through its pricing API, one offer per configuration with a stated per-GPU-hour rate. Different vCPU and memory configurations of the same GPU are separate offers at separate rates, and AIMC records each one rather than averaging them.
Nova Cloud (Nova Cloud Services Inc.) is a Canadian GPU cloud operating its own RTX 5090 and RTX PRO 6000 infrastructure in western Canada, in 1x, 2x, 4x and 8x GPU configurations, on-demand, interruptible or reserved. The company says it builds, owns and operates its servers, network and management platform, and has been building and operating GPU servers since 2018. Compute is billed by the second with no minimum usage period, interruptible instances can be displaced by on-demand or reserved ones, and reserved terms run three or six months, paid in advance.
AIMC reads Nova Cloud's published per-GPU-hour rates daily from its search API.
SimplePod is a Poland-based GPU rental platform that combines its own data center with a community cloud of hand-picked hosts, offering consumer and prosumer GPUs, including RTX 3060, RTX 3090, RTX 4070 Super, RTX 4090 and V100 SXM2 32GB, at per-GPU hourly prices. It is operated by SimplePod.ai Sp. z o.o. Sp. k., whose team launched its first product, SimpleMining.net, in 2016. GPUs are rented as Docker instances with preconfigured environments or as full virtual servers, billed with per-second accuracy only while running, and the company built its main data center in Poland with backup power and redundant fiber connections.
AIMC reads SimplePod's node-level listings daily from its public API, which publishes pricing, availability and specifications without authentication.
Lium is a GPU rental marketplace where independent providers list whole GPUs, from RTX-class cards to H100, H200, B200 and B300, at per-GPU hourly prices they set, on secure nodes or interruptible spot nodes. It describes itself as the decentralized GPU cloud on Bittensor Subnet 51: providers earn network emissions as well as rental fees, and a subnet validator regularly re-checks each node's hardware, network and container isolation, so only nodes that pass are listed. Pods are billed per second at the listed hourly price with no minimum, can be paid for by card, stablecoin or TAO, and are reached through the web console, a command-line tool, a Python SDK or the API.
AIMC reads every Lium listing daily from its public nodes feed, including the posted prices of rented nodes, with each node's location and availability as Lium states them.
ParalonCloud is a GPU rental marketplace where independent providers rent out whole machines, from RTX-class cards to the H200 NVL, at hourly prices set within ParalonCloud's catalogue range, billed by the minute, on-demand or spot. It describes itself as a decentralized GPU network whose hardware belongs to individuals and data centers around the world, each scored on its completed rentals. Spot machines are discounted but can be reclaimed by the provider at short notice, there are no contracts, minimums or egress fees, and the platform also offers an inference API and a cloud rendering service. The site is operated by NEUROSTACK SRL.
AIMC reads ParalonCloud's public offers feed daily: one price per machine shape and tier (on-demand or spot), with in-stock status as ParalonCloud states it.
Hikube is Hidora's Swiss cloud: virtual machines and managed Kubernetes with dedicated NVIDIA GPUs, on Hidora's own hardware in Switzerland, metered hourly with no minimum term. Hidora, a Swiss cloud provider active since 2016 and based in the Geneva area, introduced Hikube in 2025 as a platform that replicates data across three Swiss data centers. It is built on open-source Kubernetes, supports tools such as Terraform and Helm, and Hidora states ISO 27001 and PCI DSS certification and compliance with Swiss and EU data-protection law.
AIMC reads Hikube's published prices daily: each GPU plus the smallest instance it can run on (u1.medium), per GPU-hour, converted from Swiss francs at the ECB reference rate.
Leafcloud is a Dutch cloud provider with on-demand GPU instances in Amsterdam: A100 PCIe 80GB, A30 and RTX PRO 6000 Blackwell Server Edition, billed by the hour. Founded in 2019, it places its servers in existing buildings such as apartment complexes, care homes and swimming pools, where the heat they produce warms the building's water instead of being discarded. The platform is OpenStack-based with managed Kubernetes, all server locations are in the Netherlands, there are no data egress fees, and the company states ISO 27001 and SOC 2 Type II certifications.
AIMC reads Leafcloud's public pricing API daily: each orderable GPU instance, per GPU-hour, converted from euros at the ECB reference rate.
Enverge runs one-click GPU instances on surplus renewable energy, with regions in the United States, Brazil and Greece. On-demand instances are billed by the minute with no commitments. Its capacity sits where renewable generation is overproduced or curtailed, drawing hydro, solar and wind power at the site instead of buying offsets. When a GPU type is fully in use, new launches join a first-come queue at no charge, and larger data-center cards can be arranged through a capacity request. The service is operated by Enverge Corp.
AIMC reads Enverge's public pricing table daily: RTX PRO 6000 Blackwell Workstation Edition, B300 and DGX Spark (GB10) at the on-demand rate, per GPU-hour, in US dollars.
Micron21 is an Australian-owned and operated cloud and data centre provider. Its mCloud GPU services run in its Tier IV-certified facility in Melbourne, on its own AS38880 network with integrated DDoS protection. GPUs are dedicated PCIe cards presented to customer virtual machines via passthrough. The company began in 2004 inside an Australian printing business and describes its Melbourne facility as the country's first Tier IV data centre. mCloud is built on OpenStack and Ceph, with data synchronised across three separate data centres, and Micron21 states ISO 27001 certification and an independent IRAP assessment for government workloads.
AIMC reads Micron21's public mCloud GPU pages daily: A10, A100 PCIe 80GB, RTX PRO 6000 Blackwell and H200 NVL, each priced as a GPU with minimum specification on a 12-month term, converted from Australian dollars (ex GST) to US dollars per GPU-hour.
Dataoorts is a GPU cloud provider, founded in 2024 and headquartered in Prayagraj, India, with regions in North America, Europe and Asia. Its X-Series instances (H200 SXM5, H100, A100 80GB, L40 and others) target training and HPC, GC2 instances offer lighter VMs across a broader catalog, and prices move within a published range under its DDRA dynamic-pricing model. X-Series instances are full virtual machines on the company's in-house DDRA cluster, which it describes as combining its own GPU racks with capacity from other providers in Tier III and higher data centers. The platform also offers a serverless AI model API, prebuilt machine images and a marketplace for reserved instances.
AIMC reads Dataoorts' pricing API daily across its X-Series, Orix, Orion, Atlas, Yotta and Z-Series instances, per GPU-hour, in US dollars.
NeuralRack is a US GPU cloud, founded in 2023, operating its own bare-metal hardware in data centers in North Carolina and Florida. Its catalog features NVIDIA RTX 5090 and RTX PRO 6000 GPUs on AMD EPYC hosts with PCIe Gen 5 and 100 Gbps inter-node networking.
AIMC reads NeuralRack's published rates daily from its pricing API.
Ai Mining Co. stands out as a true leader in GPU infrastructure for its forward-thinking choice to harness prosumer RTX cards in datacenters through platforms like NeuralRack, making high-performance but reliable AI compute more accessible.
”Hydra Host is a US-based bare-metal GPU provider. Its Brokkr marketplace lists individual dedicated servers rather than a fixed catalog, each with its own GPU model, CPU, memory, storage, networking and location. The GPU models and locations on offer change from day to day, on on-demand and interruptible tiers. Customers transact in USD. Founded in 2021 and based in Miami, with further offices in Phoenix and Singapore, the company works with independent data-center operators, who keep ownership of their hardware while Hydra's software handles provisioning, billing and monitoring. Its interruptible tier offers idle servers at a discount to on-demand rates, and it also provides GPU clusters and hardware procurement services.
AIMC reads Hydra Host's Brokkr API daily. Every listing is a discrete bare-metal machine with its own specs, region and tier, and on-demand and interruptible rates are recorded separately.
Lambda is a US-based GPU cloud. Its self-serve Instances tier offers on-demand NVIDIA GPUs priced by the hour and billed in one-minute increments (B200 SXM, H100 SXM, A100 SXM 80GB and 40GB, V100), alongside 1-Click Clusters and larger reserved deployments. Founded in 2012 by AI engineers and headquartered in San Francisco, it calls itself the Superintelligence Cloud. Its 1-Click Clusters give self-serve access to multi-node GPU clusters with InfiniBand networking on short reservations, and the company also builds large dedicated deployments for training and inference at scale.
AIMC reads Lambda's headline per-GPU-hour Instances pricing daily from its public pricing page. Reserved cluster commitments are not included in these figures.
Arkane Cloud is a French GPU cloud running its own NVIDIA servers in colocation data centers in Lyon and Paris. It offers on-demand NVIDIA GPU instances priced per GPU-hour and billed by the second (B300, B200, H200, H100, A100, RTX PRO 6000, L40S, RTX 6000 Ada, RTX A6000), alongside reserved clusters and a per-token inference API. Founded in 2020 and headquartered in Lyon, it describes itself as a sovereign European AI cloud. Beyond on-demand instances with SSH access, it offers serverless GPU deployment, reserved multi-node clusters on managed Kubernetes or bare metal, and a models API serving open-source models through an OpenAI-compatible interface.
AIMC reads Arkane's on-demand per-GPU-hour rates daily from its public pricing page. Reserved clusters and per-token inference pricing are out of scope.
SynpixCloud is a GPU cloud, founded in 2024, running China-based infrastructure. Its catalog spans RTX 3080 through RTX 5090, V100, A100 SXM (40GB and 80GB), A800 SXM and H100 SXM in 1x to 8x configurations, with hourly pricing and unmetered bandwidth.
AIMC reads SynpixCloud's published product prices daily from its pricing API.
QuickPod is a US-based community GPU marketplace, founded in 2023, where individual hosts list their own hardware at prices they set. Its catalog spans RTX 5090, RTX 5070 Ti, RTX 4090, RTX 3090, A4000, A6000, RTX 6000 Ada and RTX PRO 4000 Blackwell. It describes itself as a distributed GPU and CPU marketplace: hosts set the pricing, duration and hardware grouping of their own offers, and renters launch machines from templates with tools such as Jupyter, PyTorch and TensorFlow preinstalled. The platform also supports multi-machine clusters and serverless endpoints.
AIMC reads QuickPod's listings daily through its API.
CloudRift is a GPU cloud headquartered in Santa Clara, California, offering on-demand GPU rentals, data-center management tools and serverless LLM inference. Its catalog spans NVIDIA RTX 4090, RTX 5090, RTX PRO 6000, L40S, A100, H100, H200, B200 and B300, plus AMD MI350X. Founded by Dmitry Trifonov, Slawomir Strumecki and Dimitrios Verraros, it describes itself as the operating system for sovereign AI deployments: its software lets data-center operators list, manage and rent out their own GPU capacity, and the GPUs it offers come from that network of partner data centers. Rentals are available as containers, virtual machines or bare metal at hourly rates, and the company states SOC 2 Type I certification.
AIMC reads CloudRift's on-demand and 1- and 3-month reserved rates daily from the pricing data published on its public pricing page.
OVHcloud is a European cloud provider headquartered in France. Its Public Cloud platform offers on-demand GPU instances on NVIDIA H100, A100, L40S and V100 hardware at published hourly prices. Founded in 1999 by Octave Klaba and based in Roubaix, it designs and builds its own servers and cools them with a water-cooling system it developed. The company took the OVHcloud name in 2019 and has been listed on Euronext Paris since 2021.
AIMC reads OVHcloud's GPU instance prices daily from its public order-catalog API.
Latitude.sh is a bare-metal and GPU infrastructure provider with locations across the Americas, Europe and Asia-Pacific. It offers dedicated GPU servers, including NVIDIA HGX B300 and RTX PRO 6000 configurations, with hourly prices that vary by region and monthly prices on a one-month commitment. The company began in 2001 as Maxihost, a Brazilian hosting provider, and took the Latitude.sh name in 2022 as it focused on single-tenant bare-metal servers deployed through an API. Alongside bare-metal and GPU servers, it offers virtual machines, managed Kubernetes, storage and private networking.
AIMC reads Latitude.sh's per-region hourly (no commitment) and monthly (one-month commitment) rates daily through its documented REST API.
AIME is a Berlin-based HPC system integrator and GPU cloud provider, founded in 2019, that builds and hosts the multi-GPU servers it rents. Rental configurations include H100 NVL, H200 NVL, A100 (40GB and 80GB), RTX PRO 5000/6000 Blackwell, RTX 6000 Ada, RTX 3090 and RTX 4090, in 1x to 8x configurations on weekly, monthly, half-yearly and yearly terms. Founded by Toine Diepstraten and Henri Hagenow, AIME GmbH designs and assembles its own HPC servers and workstations, and customers can also buy hardware and have it hosted in its Berlin colocation. Rented servers come with Ubuntu and its AIME MLC container framework preinstalled, and the company states that its cloud hosting and offices run entirely on electricity from renewable sources.
AIME publishes its cloud prices as a public JSON file, and AIMC reads that file daily, publishing the monthly-plan rate per GPU-hour. AIME rents on fixed terms (weekly and longer), not hourly on demand.
Qubrid AI is an AI cloud platform based in McLean, Virginia, offering on-demand GPU virtual machines, bare-metal servers, multi-node clusters and serverless inference. Its on-demand lineup includes B200, H200, H100, A100 (40GB and 80GB), L40S, A10G, L4 and T4 in 1x, 4x and 8x configurations, billed per second. Founded by Pranay Prakash, it describes itself as a full AI stack and lists NVIDIA as an official partner. Beyond on-demand virtual machines, it offers a serverless model API for open-source models, reserved enterprise GPU clusters, AI appliances for on-site use and managed hosting of customers' GPU infrastructure.
AIMC reads Qubrid's on-demand GPU rates daily from the instance data published on its public pricing page. Serverless inference pricing is out of scope.
Cyfuture AI is an AI cloud platform headquartered in India, with data centers in Noida, Jaipur, Raipur and Bangalore, offering GPU-as-a-Service and clusters alongside inference, fine-tuning and other AI services. Its on-demand GPU lineup includes NVIDIA H100 SXM, A100, L40S and V100, AMD MI300X and MI325X, and Intel Gaudi2, in 1x to 8x configurations billed per second in INR. It is the AI arm of Cyfuture India Pvt. Ltd., an IT services company operating since 2001, and runs its services from the group's own Tier III data centers. Cyfuture states MeitY empanelment along with ISO 27001, PCI DSS and SOC certifications, and says customer data stays within India.
AIMC reads Cyfuture's on-demand rates daily from its public pricing page and converts INR to USD at the ECB reference rate.
Packet.ai is an on-demand GPU cloud from hosted.ai (HOSTED AI Inc.), with infrastructure based in Sweden. Its catalog includes NVIDIA RTX PRO 6000 Blackwell, A100 80GB and B200, run as isolated containers with dedicated GPU memory and no preemption. It offers two tiers: Dynamic GPUs, which are pooled cards shared between isolated tenants at a lower price, and Dedicated GPUs, where a full card is reserved for one customer, which is the tier AIMC observes. Usage is billed by the minute from a prepaid wallet with no contract or minimum, and instances come with root access over SSH, a web terminal or Jupyter.
AIMC reads Packet.ai's dedicated (single-tenant) on-demand rates daily from its public pricing page.
CloudPe is an Indian cloud infrastructure provider, founded in 2024 and headquartered in Pune, with data-center capacity in Mumbai. Built on OpenStack and operated in partnership with Leapswitch, it offers NVIDIA H200, L4 and RTX PRO 6000 GPUs. It describes itself as a joint venture of Leapswitch Networks, a long-established hosting provider, and Strad Solutions, a managed-services company. Alongside GPU instances, its platform offers virtual machines, managed Kubernetes, private networking and object storage.
AIMC reads CloudPe's GPU prices daily from the structured data (Schema.org) published on its GPU product pages, and converts INR to USD at the ECB reference rate.
DeepInfra is a US-based machine-learning infrastructure provider, founded in 2022, offering both serverless inference APIs and dedicated GPU rentals. Its dedicated catalog includes NVIDIA A100, H100, H200, B200 and B300 billed by the minute, alongside multi-node clusters. Founded by Nikola Borisov, Georgios Papoutsis and Yessenzhar Kanapin, it describes itself as inference infrastructure built from the GPU hardware up to the API layer. Its GPU Instances launch as containers reached over SSH, with no minimum commitment and no egress fees.
AIMC reads DeepInfra's dedicated GPU rates daily through its API. Per-token inference pricing is out of scope.
Thunder Compute is a US-based on-demand GPU cloud. Its catalog includes NVIDIA A100 80GB, H100 SXM, L40 and RTX A6000, in configurations of up to 8 GPUs, billed by the minute and never preempted. Founded by Carl Peterson and Brian Model, based in San Francisco and backed by Y Combinator, it calls itself the GPU sandbox company: alongside persistent GPU instances for developers, it offers GPU sandboxes for AI agents and automated testing. Instances can be managed from a VS Code extension, a command-line tool or its API, and disk snapshots let users delete an instance and restore it later.
AIMC reads Thunder Compute's rates daily from its published pricing API. The production and prototyping tiers are distinct products.
Together AI is a US-based AI cloud offering self-service NVIDIA HGX H100, H200, B200 and B300 clusters with InfiniBand networking, on-demand, preemptible or reserved, alongside its inference platform. Founded in 2022 by a team including Vipul Ved Prakash and based in San Francisco, the company is best known for its platform for running, fine-tuning and training open-source generative AI models. Its GPU clusters are created from the console or command line, run on Kubernetes with optional Slurm, and can mix on-demand capacity billed hourly with reservations of 3 to 180 days, with no data ingress or egress fees.
AIMC reads Together AI's on-demand and preemptible per-GPU-hour cluster rates daily from its public pricing page.
Theta EdgeCloud is the GPU compute platform from Theta Labs, founded in 2018. It combines managed cloud shards on NVIDIA H100, H200, A100, V100 and T4 GPUs with a community tier where independent node operators contribute consumer and workstation GPUs to a peer-to-peer rental pool. Customers transact in USD; TFUEL settles between Theta and node operators. EdgeCloud launched in May 2024 as a hybrid platform, pairing data-center GPUs from cloud partners with Theta's network of community-run edge nodes and using the Theta blockchain to register and verify computing tasks. It is used for AI model deployment, video processing and rendering, with models launched from templates or developed in Jupyter notebooks.
AIMC reads Theta EdgeCloud's GPU prices daily through its API.
UpCloud is a Finnish cloud provider, founded in 2011 and headquartered in Helsinki, offering on-demand NVIDIA L4, L40S, H100 and B200 GPU servers from its Helsinki (fi-hel2) data center, with one to eight dedicated GPUs per server and hourly billing. Founded by Joel Pihlajamaa, who remains its chief executive, it describes itself as a European cloud provider with data centers on four continents, most of them running on renewable energy. Alongside on-demand GPU servers, it offers dedicated NVIDIA GPU clusters for larger deployments.
AIMC reads UpCloud's hourly GPU server prices daily from its public pricing page. UpCloud's API does not publish prices.
Identical data collection and presentation for every marketplace. No partner receives preferential treatment.
Daily rental pricing across all GPU models tracked on your marketplace.
Full normalization of GPU names and variants across marketplaces for accurate comparison.
How many GPUs are available at any given time, broken down by model.
Datacenter locations and regional distribution of GPU inventory.
Spot, on-demand, and reserved pricing separated and labeled clearly.
VRAM, interconnect type, and generation details for every tracked GPU.
UTC-timestamped data collection every 24 hours, no manual intervention.
Equal-placement profile on our platform with your branding and live stats.
From first contact to live data. Same process for every marketplace.
Reach out to discuss your marketplace. We evaluate fit based on data quality and API access — not payment.
We assess API availability, data formats, rate limits, and authentication to build a reliable integration.
We verify pricing accuracy, listing counts, and GPU model normalization against live data.
Your marketplace appears with full pricing coverage, live stats, and equal showcase placement.
Every marketplace receives identical showcase placement, data coverage, and visibility. No preferred positioning. No exceptions.
We only collect data that each marketplace has authorized in writing. No unauthorized access. No gray areas.
We evaluate partnerships based on data quality, API reliability, and marketplace activity. Every partner meets the same standards.
We evaluate every marketplace using the same data quality and integration standards.
Equal treatment. One standard. If your marketplace publishes GPU rental prices, we'd like to hear from you.
Ai Mining Co. is a product of AIMC Technologies, LLC