Vultr GPU Cloud Review 2026
Vultr offers GPU cloud instances across 32 global locations with A100s starting at $0.65/hr. A solid mid-tier option with global reach. Here's our full review.
Overview
Vultr is a global cloud infrastructure provider with 32 data centre locations. They've expanded into GPU cloud with NVIDIA A100 and H100 instances, targeting ML inference, fine-tuning, and AI application deployment. Their strength is geographic coverage — you can deploy GPUs closer to your users than with most GPU specialists.
Vultr's GPU offering sits between the budget marketplaces (Vast.ai) and hyperscalers (AWS/GCP), offering a balance of reliability, global reach, and competitive pricing.
Pricing (March 2026)
| Instance | GPU | VRAM | vCPUs | RAM | Price |
|---|---|---|---|---|---|
| vcg-a100-1c-80g | 1x A100 | 80 GB | 12 | 120 GB | $0.65/hr |
| vcg-a100-2c-160g | 2x A100 | 160 GB | 24 | 240 GB | $1.30/hr |
| vcg-a100-4c-320g | 4x A100 | 320 GB | 48 | 480 GB | $2.60/hr |
| vcg-h100-1c-80g | 1x H100 | 80 GB | 24 | 240 GB | $2.88/hr |
| vcg-h100-2c-160g | 2x H100 | 160 GB | 48 | 480 GB | $5.76/hr |
| vcg-h100-8c-640g | 8x H100 | 640 GB | 192 | 1920 GB | $23.04/hr |
| vcg-l40s-1c-48g | 1x L40S | 48 GB | 12 | 120 GB | $0.95/hr |
All instances include local NVMe SSD storage and high-speed networking. Billed hourly with no minimum commitment. Reserved instances available for 12-month terms at ~30% discount.
Key Features
- 32 Global Locations: Deploy GPUs in North America, Europe, Asia, and Australia.
- Kubernetes Integration: Vultr Kubernetes Engine supports GPU node pools.
- Marketplace: One-click ML app deployment (Ollama, vLLM, text-generation-inference).
- Block Storage: Attach up to 40TB of block storage per instance.
- Full API: REST API for programmatic instance management.
- Reserved Instances: Save ~30% with 12-month commitments.
Pros
- A100 at $0.65/hr — excellent value
- 32 global locations for low-latency inference
- Kubernetes-native GPU support
- Reserved instance discounts available
- Simple, predictable pricing
- Good uptime and reliability
Cons
- Limited GPU variety (mainly A100/H100/L40S)
- No spot/interruptible pricing
- H100 pricing not as competitive ($2.88/hr)
- No serverless inference platform
- GPU availability varies by region
- Less ML-focused tooling than specialists
Verdict: 3.9/5
Vultr is an excellent choice for inference workloads that need global distribution. The A100 at $0.65/hr is very competitive, and 32 locations mean you can deploy close to users worldwide. Less ideal for large-scale training where marketplace providers offer better prices on H100 clusters.
Who Should Use Vultr GPUs?
- Global inference: Deploy models in 32 locations for low-latency serving
- A100 workloads: Best-in-class A100 pricing at $0.65/hr
- Kubernetes teams: Native GPU support in VKE
Try Vultr GPU Cloud
A100 from $0.65/hr. Deploy across 32 global locations.
Get Started on Vultr →Perffeco may earn a commission from this link at no extra cost to you.