deploybase

Vultr GPU Cloud Pricing: Complete Guide for Every GPU (March 2026)

Deploybase · November 25, 2025 · GPU Pricing

Contents

Vultr GPU Pricing Overview

Vultr offers GPU cloud compute as an alternative to major cloud providers. The platform targets developers and small teams seeking simplicity without production complexity.

As of March 2026, Vultr pricing sits between budget providers (Vast.AI) and production clouds (AWS). Reliability is good but less mature than established platforms.

This guide breaks down Vultr GPU offerings, pricing, and cost strategies.

Vultr's Market Position

Vultr competes on:

  • Simplified pricing (no hidden fees)
  • Global data center presence
  • Transparent billing
  • Good uptime (99.99% SLA in documentation)

Vultr is smaller than AWS, Azure, Google Cloud. It's larger and more reliable than peer-to-peer markets like Vast.AI.

GPU Instance Pricing

H100 GPU Instances (Bare Metal)

8×H100 bare metal:

  • Hourly: $23.92/hr (full bare metal server, 8 GPUs)
  • Per-GPU equivalent: ~$2.99/hr
  • Best for: Large distributed training requiring dedicated hardware

NVIDIA GH200 Instances

Single GH200:

  • Hourly: $1.99/hr
  • Best for: Large model inference with massive unified memory (141GB HBM3e + 480GB LPDDR5X)

GH200 at $1.99/hr is exceptionally cost-effective for memory-bound workloads. The Grace Hopper Superchip integrates GPU and CPU with high-bandwidth interconnect.

Comparison for GH200:

  • Vultr GH200: $1.99/hr
  • DigitalOcean MI300X: $1.99/hr (similar price, AMD alternative)

A100 PCIe GPU Instances

Single A100 PCIe:

  • Hourly: ~$2.397/hr
  • Memory: 80GB RAM
  • vCPUs: 24 cores
  • Storage: 50GB SSD

Comparison at $2.397/hr for A100 PCIe:

  • Vast.AI: $0.80-1.50/hr (variable marketplace)
  • RunPod: $1.39/hr (cheaper)
  • Lambda A100: $1.48/hr (cheaper)
  • AWS (per GPU): ~$3.00-4.00/hr (p4d instances, more expensive)
  • Google Cloud: ~$3.67/hr (a2-highgpu-1g on-demand, more expensive)

Vultr A100 is competitive versus major clouds.

T4 GPU Instances

T4 single GPU:

  • Monthly: ~$0.90/hr
  • Annual: ~$0.68/hr
  • Memory: 32GB RAM
  • vCPUs: 12 cores

Budget option. Slower but cheaper. Suitable for small inference or testing.

Cost comparison:

  • Vast.AI T4: ~$0.15-0.30/hr
  • Vultr: $0.90/hr
  • AWS T4: ~$0.35/hr

Vast.AI is 3-6x cheaper for T4.

Multi-GPU Instances

Vultr offers multi-GPU configurations but pricing scales linearly. Two T4s cost 2x one T4. No bulk discounts.

Regional Pricing

Vultr has 24+ global data centers. Pricing varies:

Lowest cost regions:

  • New Jersey: baseline pricing
  • Los Angeles: +5-10%
  • London: +15-20%
  • Singapore: +25-30%
  • Sydney: +30-35%

US regions are cheapest. Asia-Pacific costs more. Pick region based on latency needs, not cost (differences are small).

Committed Discounts

Annual Commitment

Vultr offers ~25% discount for annual prepayment:

A100 PCIe example:

  • Hourly: $2.397/hr = $20,998/year
  • Annual commitment: ~$1.80/hr = $15,768/year (est. 25% discount)
  • Savings: ~$5,230/year (25%)

Significant for committed workloads.

Monthly vs Hourly

No hourly vs monthly distinction on Vultr. Billing is hourly with annual prepayment discounts.

Vultr vs Competitors

A100 PCIe pricing (single GPU):

  • Vultr: $2.397/hr
  • Vast.AI: $0.80-1.50/hr (variable marketplace, can be cheaper)
  • RunPod: $1.39/hr (cheaper)
  • Lambda A100: $1.48/hr (cheaper)
  • AWS: ~$3.00-4.00/hr per GPU (p4d instances, more expensive)
  • Google Cloud: ~$3.67/hr (more expensive)

GH200 pricing:

  • Vultr: $1.99/hr
  • DigitalOcean MI300X: $1.99/hr (similar)

H100 bare metal (8-GPU):

  • Vultr: $23.92/hr total ($2.99/GPU equivalent)
  • Lambda H100 SXM: $3.78/hr (cheaper per GPU)
  • CoreWeave 8×H100: $49.24/hr (more expensive)

T4 pricing:

  • Vultr: $0.90/hr
  • Vast.AI: $0.15-0.30/hr (3-6x cheaper)
  • AWS: $0.35/hr (cheaper)
  • Google Cloud g2: $0.35/hr (cheaper)

Why Choose Vultr Despite Higher Cost?

Simplicity: No provider selection like Vast.AI. No commitment lock-in like AWS.

Reliability: Better than Vast.AI (peer-to-peer), similar to AWS.

Support: Human support team. Better than peer-to-peer.

Compliance: Meets certain certifications. Useful for regulated industries.

For pure cost, Vast.AI wins. For simplicity, Vultr is reasonable despite higher cost.

Cost Optimization Strategies

1. Use Annual Commits

25% discount for annual prepayment. If budget allows, commit upfront.

2. Right-size instances

Don't rent A100 instances for small models. Use T4 instead. Cost difference is 3x.

3. Stop instances when idle

Stopped instances don't charge for compute. Storage charges continue. Stop when not training.

4. Choose US regions

Cheapest pricing. No latency penalty unless serving globally.

5. Monitor usage

Vultr's billing dashboard is straightforward. Set up alerts for budget overruns.

6. Batch processing

Process multiple jobs sequentially. Reduces per-job startup overhead.

7. Combine Vultr with Vast.AI

Use Vultr for reliability-critical work. Use Vast.AI for budget-critical work. Distribute workloads accordingly.

Storage and Bandwidth

Vultr charges separately for:

Storage: $0.10/GB/month (~$100/TB/year)

  • Higher than Google Cloud ($20/TB)
  • Similar to AWS

Bandwidth: $0.01/GB outbound

  • Cheaper than AWS ($0.09/GB to internet)
  • More expensive than Google Cloud ($0.12/GB but includes more allowance)

Monitor data transfer. Downloading large datasets costs money.

Performance Characteristics

Vultr's GPU instances are:

  • Older hardware (A100s from 2021 batches, for example)
  • Shared infrastructure (not dedicated)
  • Predictable performance (not spot pricing)

Older hardware is slower than newest generation but still current for production work. Sharing infrastructure means performance varies slightly.

Total Cost of Ownership Example

Monthly A100 PCIe usage (24/7 for 1 month):

  • Compute: $2.397 × 730 = $1,750
  • Storage (500GB): $0.10 × 500 × 1 = $50
  • Bandwidth (100GB): $0.01 × 100 = $1
  • Total: ~$1,801

Compare to Vast.AI:

  • Compute: $1.00 × 730 = $730 (average)
  • Storage: $0.05 × 500 = $25
  • Bandwidth: Free
  • Total: ~$755

Vultr is ~2.4x more expensive than Vast.AI average for A100, but offers guaranteed reliability.

When Vultr Makes Sense

Use Vultr when:

  • Developers want simplicity over absolute lowest cost
  • Reliability is critical (more reliable than Vast.AI)
  • Developers need support (human support team)
  • Developers have compliance requirements
  • Long-term commitment is acceptable (annual prepay)

Avoid Vultr when:

  • Cost is paramount
  • Developers need flexibility (Vast.AI is more flexible)
  • latest hardware matters (developers get older generations)
  • Developers need spot pricing (Vultr doesn't offer it)

Vultr Free Trial

Vultr offers $250 free credit for new users. Covers a few days of T4 or A100 testing. Try before committing.

FAQ

Is Vultr cheaper than AWS for GPU?

No. Vultr is more expensive for single GPUs but might be cheaper for multi-GPU instances. AWS is generally better value.

Should I use Vultr or Vast.AI?

Use Vast.AI if cost matters. Use Vultr if simplicity and reliability matter. Vast.AI is 3-5x cheaper but more complex.

Can I run production ML on Vultr?

Yes. Reliability is good. Support is available. Cost is high but acceptable for production if budget allows.

Do Vultr instances get interrupted?

No. Vultr is not spot. Instances are committed. Outages are rare.

What's the best GPU to start with on Vultr?

T4 at $0.90/hr for testing. GH200 at $1.99/hr for memory-intensive inference. A100 PCIe at $2.397/hr for standard training. 8×H100 bare metal at $23.92/hr for large-scale distributed training.

How do I estimate my Vultr bill?

(GPU hours × hourly rate) + storage + bandwidth. Vultr pricing calculator is available on website.

Can I downsize or upgrade instances?

Yes. Stop instance and resize. CPU, memory, and storage changes available. GPU type changes require new instance.

Sources

  • Vultr GPU Pricing (as of March 2026)
  • Vultr Instance Type Documentation
  • Performance Benchmarking Data
  • Cloud Pricing Comparisons (March 2026)

Last updated: March 2026. Pricing reflects market rates as of March 22, 2026.