GPU Cloud Providers

Google's cloud with strong ML tooling via Vertex AI and TPU access. Offers A100, H100, L4, and T4 GPUs on a2, a3, and g2 machine types. Best sustained-use discounts in the industry.

The world's largest cloud provider with the broadest GPU selection. AWS offers P4d, P5, and G-series instances powered by NVIDIA A100, H100, and L4 GPUs across 30+ regions. Enterprise-grade with full SLA coverage.

Microsoft's cloud platform with deep OpenAI integration. Azure NC, ND, and NV-series VMs offer A100, H100, and A10 GPUs. Preferred by enterprises already in the Microsoft ecosystem.

Oracle Cloud offers bare-metal A100 and H100 GPU instances at highly competitive pricing. Known for the cheapest egress in the industry at $0.009/GB and aggressive pricing to win ML workloads.

IBM's enterprise cloud platform offering GPU virtual servers with V100 and A100 GPUs. Full enterprise compliance suite including FedRAMP. Best for regulated industries.

Asia's largest cloud provider offering A100, V100, and T4 GPU instances across APAC, EU, and US regions. Alibaba Cloud is the go-to option for teams serving Asian markets.

Tencent's cloud platform offering GPU instances for AI/ML across Asia-Pacific regions. Offers V100, A100, and T4 GPUs with competitive APAC pricing.

Developer-friendly GPU cloud with the widest GPU selection in the market. RunPod offers spot and on-demand pricing across consumer and data center GPUs, plus serverless inference endpoints.

GPU aggregator providing a single API to provision GPUs from multiple underlying cloud providers. Shadeform finds the cheapest available GPU across its partner network.

ML-focused GPU cloud built by AI researchers. Lambda offers H100, A100, and A10 instances with simple pricing — storage and egress are included. Popular with startups and research labs.

General-purpose cloud provider with GPU Cloud and GPU Bare Metal offerings. Vultr provides A100, H100, and L40S GPUs across 30+ global locations with simple hourly billing.

Budget GPU cloud with some of the lowest H100 prices on the market. Verda targets cost-conscious teams with aggressive pricing on NVIDIA GPUs.

EU-based GPU cloud spun out of Yandex Cloud. Nebius offers H100 and L40S instances with competitive pricing and low egress costs. Good option for teams needing European data residency.

Specialized GPU cloud focused on AI workloads. CloudRift offers competitive pricing on NVIDIA GPUs with a developer-focused experience.

Finland-based ML-focused GPU cloud offering A100, H100, and L40S instances. Datacrunch provides spot and on-demand pricing with free egress and dedicated ML tooling.

French cloud provider with strong EU data sovereignty positioning. Scaleway offers H100, L4, and RENDER-S GPU instances across Paris and Amsterdam. GDPR and ISO 27001 compliant.

Major European cloud with one of the largest data center networks in Europe. OVHcloud offers A100 and V100 GPUs in both VM and bare metal configurations. HDS-certified for healthcare workloads.

UK-based NVIDIA partner cloud offering H100, A100, and L40S GPUs. Hyperstack focuses on competitive pricing and fast provisioning for AI/ML workloads.

US-based GPU cloud powered by otherwise-wasted stranded energy (flared gas). Crusoe offers competitive H100 pricing with 200 Gbps networking, making it compelling for sustainable AI training.

Budget-friendly ML platform popular with indie AI developers and students. Jarvis Labs offers A100, A6000, and RTX GPUs with pre-installed ML frameworks and persistent storage.

Akamai-owned cloud provider with GPU Linodes featuring RTX 6000 Ada GPUs. Known for developer-friendly docs and simple pricing. Good option for inference and smaller training jobs.

NVIDIA-backed GPU cloud purpose-built for AI/ML at scale. CoreWeave offers H100 SXM clusters with 400 Gbps InfiniBand networking, making it one of the top choices for large-scale training workloads.

ML development platform from the creators of PyTorch Lightning. Lightning.ai offers GPU-powered "Studios" with pre-built environments for training, fine-tuning, and experimentation.

Iceland-based GPU cloud powered by 100% renewable energy. Genesis Cloud offers GPUs with storage and egress included, making it simple and green. ISO 27001 certified.

API-first GPU cloud aggregating distributed compute resources. FluidStack offers competitive pricing on H100, A100, and A6000 GPUs with a developer-friendly API.

Global edge cloud provider offering GPU instances powered by L40S and A100 GPUs. Gcore operates data centers across Europe, Asia, and Americas with ISO 27001 certification.

India's leading AI/GPU cloud provider, publicly listed on NSE. E2E Networks offers H100, A100, and A30 GPUs from Indian data centers with competitive pricing.

European GPU rental service offering dedicated GPU servers for AI/ML and rendering workloads. LeaderGPU provides multi-GPU configurations in EU data centers.

ML platform acquired by DigitalOcean, popular for Gradient notebooks and GPU VMs. Paperspace offers A100, A6000, and RTX GPUs with a clean developer experience and persistent storage.

Fast-growing GPU cloud platform offering competitive pricing on H100 and A100 instances. Popular for AI training and inference workloads.

Budget GPU cloud offering both Linux and Windows GPU VMs. Popular for both AI/ML and creative workloads like 3D rendering.

French GPU cloud provider offering sovereign EU infrastructure. Competitive H100 pricing with GDPR-compliant data residency for European enterprises.

German budget hosting provider offering affordable GPU VPS and dedicated servers. Good option for non-critical workloads where cost is the primary concern.

100% renewable energy GPU cloud based in Scandinavia. Offers H100 and A100 clusters powered by hydroelectric energy with GDPR compliance.

Enterprise-grade GPU cloud based in San Diego, CA. Offers H100 and A100 clusters with SOC 2 compliance for regulated industries.

Budget GPU cloud provider offering affordable access to NVIDIA GPUs for AI/ML workloads.

European sovereign GPU cloud focused on AI sovereignty and GDPR compliance. Nscale offers H100 and A100 GPUs from EU data centers.

UK-based cloud platform with Kubernetes-native GPU instances. Popular with developers for its simplicity, fast provisioning, and competitive EU pricing.

Developer-focused platform with GPU VMs deployable globally via anycast routing. Popular for ML inference workloads with its CLI-driven workflow and fast provisioning.

US-based GPU cloud specializing in AMD MI300X and H100 instances. Offers one of the largest MI300X fleets with competitive pricing for training workloads.

Budget GPU cloud provider targeting cost-conscious AI developers. Offers competitive pricing on H100 and A100 instances.

Developer-friendly cloud known for simplicity. DigitalOcean's GPU Droplets offer H100 SXM GPUs with the same clean UX developers love. SOC 2 and ISO 27001 compliant.

US-based GPU cloud specializing in H100 clusters for AI training. Voltage Park offers competitive pricing on high-end NVIDIA hardware.

Emerging GPU cloud provider targeting AI/ML workloads with competitive pricing on high-end GPUs.

Peer-to-peer GPU marketplace connecting renters with individual hosts. Vast.ai offers the cheapest GPUs available — often 3-5x cheaper than cloud providers — but with variable reliability and ephemeral storage.

GPU marketplace with flexible instance configurations. TensorDock lets you customize CPU, RAM, and GPU combinations at budget prices. Good for experimentation and development workloads.

Distributed GPU network running on consumer hardware worldwide. Salad Cloud offers ultra-cheap GPU compute for batch inference, image generation, and other parallelizable workloads.

Decentralized GPU compute platform aggregating compute from data centers worldwide. Cudo Compute offers competitive pricing on A100 and consumer GPUs with a focus on sustainable energy.

Serverless GPU platform for running Python functions in the cloud. Modal bills per second with zero idle costs — deploy inference endpoints or run batch jobs without managing infrastructure.

Serverless GPU platform with a Python SDK for deploying ML workloads. Per-second billing with automatic scaling and no infrastructure management required.

Serverless ML infrastructure platform offering per-second GPU billing with a Python-first API. Deploy models in minutes without managing infrastructure.

Model hosting platform that lets you run open-source ML models in the cloud with a single API call. Replicate handles infrastructure while you pay per second of GPU time.

Serverless GPU platform headquartered in France with GDPR-compliant EU infrastructure. Offers per-second billing and a developer-friendly deployment model.

AI platform offering inference APIs and dedicated GPU clusters for fine-tuning. Together AI provides access to open-source models via API and dedicated GPU training on H100 clusters.

Serverless GPU inference platform optimized for image generation and media AI. Per-millisecond billing with fast cold starts and a marketplace of pre-built models.

High-performance inference platform specializing in fast model serving. Fireworks AI offers serverless inference for open-source LLMs with industry-leading latency.

Major German hosting provider known for unbeatable price-performance. Hetzner offers dedicated GPU servers with storage and bandwidth included — no hidden costs. ISO 27001 and GDPR compliant.

Lithuanian bare metal hosting provider offering dedicated GPU servers with NVIDIA GPUs. Cherry Servers provides ISO 27001 certified infrastructure in EU data centers.

Bare metal GPU cloud with global locations. Latitude.sh offers dedicated NVIDIA GPU servers with no virtualization overhead, storage and bandwidth included in the price.