Cloud and inference providers
Understand delivery modes and prepare comparable requirements for a potential supplier.
Computraq is preparing for launch. These guides explain readiness and the proposed framework; they do not establish a live service, integration or partnership.
24 guides
- Amazon Web Services: what to verify before choosing
AWS provides GPU and other accelerated EC2 instance families; orchestration and managed model services are separate layers. Prepare a workload-specific quote and acceptance check.
- Baseten: what to verify before choosing
Baseten packages and deploys models as managed GPU inference endpoints, with Truss workflows and autoscaling controls. Prepare a workload-specific quote and acceptance check.
- CoreWeave: what to verify before choosing
Serverless, dedicated and self-managed Kubernetes inference each shift a different amount of GPU operations to the provider or customer. Prepare a workload-specific quote and acceptance check.
- Crusoe: what to verify before choosing
Crusoe offers customer-managed GPU infrastructure alongside managed Kubernetes, Slurm and separate managed inference paths. Prepare a workload-specific quote and acceptance check.
- CUDO Compute: what to verify before choosing
CUDO Compute now centers on dedicated GPU clusters and managed AI infrastructure; its official status page says the public on-demand platform closed March 31, 2026.
- Fireworks AI: what to verify before choosing
Fireworks AI distinguishes token-based serverless APIs, dedicated GPU deployments and reserved capacity for model serving. Prepare a workload-specific quote and acceptance check.
- Fluidstack: what to verify before choosing
Fluidstack’s public docs describe bare-metal managed Kubernetes and Slurm clusters for inference and distributed training. Prepare a workload-specific quote and acceptance check.
- GMI Cloud: what to verify before choosing
GMI Cloud spans serverless inference APIs, dedicated GPU endpoints and infrastructure clusters for training and serving. Prepare a workload-specific quote and acceptance check.
- Google Cloud: what to verify before choosing
Google Cloud offers GPU VMs, GKE GPU nodes and a distinct TPU path with different software and workload requirements. Prepare a workload-specific quote and acceptance check.
- Hugging Face Inference Endpoints: what to verify before choosing
Hugging Face Inference Endpoints deploy Hub models on managed dedicated infrastructure with selectable engines and autoscaling. Prepare a workload-specific quote and acceptance check.
- Hyperstack: what to verify before choosing
Hyperstack offers GPU virtual machines and Kubernetes clusters alongside a separate OpenAI-compatible AI Studio service. Prepare a workload-specific quote and acceptance check.
- Lambda: what to verify before choosing
Lambda separates on-demand GPU VMs, reserved multi-node clusters and single-tenant Private Cloud capacity. Prepare a workload-specific quote and acceptance check.
- Microsoft Azure: what to verify before choosing
Azure offers workload-specific GPU VM families, with AKS and Azure Machine Learning as separate orchestration services. Prepare a workload-specific quote and acceptance check.
- Modal: what to verify before choosing
Modal runs code-defined Python functions and GPU workloads on serverless compute that scales with invocation demand. Prepare a workload-specific quote and acceptance check.
- Nebius: what to verify before choosing
GPU virtual machines and clusters pair direct workload control with Nebius AI Cloud orchestration and platform services. Prepare a workload-specific quote and acceptance check.
- Nscale: what to verify before choosing
Nscale describes managed serverless or dedicated inference on its AI infrastructure, with service boundaries to confirm per offer. Prepare a workload-specific quote and acceptance check.
- NVIDIA DGX Cloud: what to verify before choosing
DGX Cloud is NVIDIA’s internal AI environment plus configuration-specific offers through cloud platforms and partners. Prepare a workload-specific quote and acceptance check.
- Replicate: what to verify before choosing
Replicate runs hosted model APIs and custom deployments, while managing model execution, scaling and prediction lifecycle. Prepare a workload-specific quote and acceptance check.
- Runpod: what to verify before choosing
Runpod separates directly managed GPU Pods from request-driven Serverless endpoints with managed worker lifecycle. Prepare a workload-specific quote and acceptance check.
- Sakura Internet: what to verify before choosing
Sakura’s Koukaryoku family distinguishes dedicated PHY servers, VRT GPU VMs and DOK container task execution. Prepare a workload-specific quote and acceptance check.
- Scaleway: what to verify before choosing
Scaleway pairs GPU virtual instances for direct control with Managed Inference for hosted model endpoints. Prepare a workload-specific quote and acceptance check.
- Together AI: what to verify before choosing
Together AI spans serverless model APIs, dedicated inference endpoints and GPU clusters with increasing infrastructure control. Prepare a workload-specific quote and acceptance check.
- Vast.ai: what to verify before choosing
Vast.ai combines a marketplace of host-provided GPU instances with on-demand, reserved, interruptible and serverless options. Prepare a workload-specific quote and acceptance check.
- Verda: what to verify before choosing
Verda covers GPU instances, virtualized Instant Clusters and autoscaling Serverless Containers, with custom bare metal by request. Prepare a workload-specific quote and acceptance check.
No matching guides.