Updated June 9, 2026 — live data from corvex.io

Corvex GPU Cloud
Pricing 2026

Corvex is a GPU-native cloud provider built on CoreWeave infrastructure — a CoreWeave subsidiary. B300 at $4.48/hr is the cheapest single-GPU rate in GridStackHub's entire 34-provider network. H100 reserved 1yr from $1.59/hr.

$4.48/hr
B300 on-demand
Cheapest in network
$1.59/hr
H100 reserved 1yr
Best committed rate
$3.15/hr
H100 on-demand
vs $2.99 network avg
6
GPU models
H100/H200/A100/B200/B300

🏆 Network's Cheapest B300: $4.48/hr

Corvex's B300 at $4.48/hr beats every provider in GridStackHub's 34-provider network — CoreWeave ($5.20), Lambda ($4.95), Vast.ai ($5.49), RunPod ($6.94), TensorDock ($4.99), Shadeform ($4.95). The 192GB HBM3e B300 at $4.48/hr is a step-change in Blackwell value.

For memory-bound inference on 70B+ parameter models, the B300 at $4.48/hr is now the go-to choice in GridStackHub's tracking network.

$4.48
/hr per GPU
B300 SXM · 192GB HBM3e
−27% vs network avg of $6.12
GPU-Native Cloud

About Corvex

Corvex is a GPU-native cloud provider built on CoreWeave's infrastructure — effectively a subsidiary or sibling to CoreWeave with a focus on competitive pricing. It offers GPU compute across the full Blackwell and Hopper stack, with particular strength in reserved pricing (1-year H100 at $1.59/hr is among the best reserved rates in GridStackHub's network).

  • B300 at $4.48/hr — cheapest single-GPU in network
  • H100 1-year reserved at $1.59/hr — strong commitment value
  • CoreWeave infrastructure backbone — proven hardware
  • Blackwell-first (B200, B300 available now)
  • InfiniBand interconnect for multi-GPU cluster workloads

Ideal Use Cases

Corvex is strongest for teams who want Blackwell access without paying a premium, and for organizations willing to commit to 1-year reserved contracts for maximum savings.

  • Large-scale LLM inference on B300 (70B+ param models)
  • Training workloads with 1-year horizon (reserved H100)
  • Enterprise AI infrastructure with committed spend
  • Distributed training requiring InfiniBand interconnect
  • Teams migrating from CoreWeave seeking better rates

💰Corvex GPU Pricing — Live Rates

Live data — updated June 9, 2026 · Source: corvex.io/gpu-pricing
GPU Model VRAM Instance Type Hourly Rate Pricing Type vs Network Avg Status
B300 SXM 192 GB HBM3e 1x B300 SXM $4.48/hr On-demand −27% vs avg Blackwell · Best
B200 SXM 192 GB HBM3e 1x B200 SXM $6.99/hr On-demand −13% vs avg Blackwell
H200 SXM 141 GB HBM3e 1x H200 SXM $3.79/hr On-demand −7% vs avg Verified
H100 SXM5 80 GB HBM3 1x H100 SXM $3.15/hr On-demand +5% vs avg Verified
H100 SXM5 80 GB HBM3 1x H100 SXM (1yr) $1.59/hr Reserved 1yr −47% vs avg Best reserved
A100 SXM 80GB 80 GB HBM3e 1x A100 SXM 80GB $2.19/hr On-demand ±0% vs avg Verified

GridStackHub network average: B300 $6.12/hr · B200 $8.60/hr · H200 $4.36/hr · H100 $2.99/hr · A100-80GB $2.19/hr. Corvex beats the network on 5 of 6 tracked GPU models.

⚡Corvex GPU Specifications

B300
192GB HBM3e · NVLink 5
B200
192GB HBM3e · NVLink 5
H200
141GB HBM3e · InfiniBand
H100
80GB HBM3 · InfiniBand
A100-80GB
80GB HBM3e
Interconnect
InfiniBand (cluster)
Best Reserved
H100 $1.59/hr (1yr)
Infrastructure
CoreWeave backbone

📊Corvex B300 vs All Providers — Cheapest B300 Comparison

B300 on-demand hourly rate across GridStackHub's provider network. Sorted by price.

Corvex
$4.48
B300 SXM on-demand
✓ Best
Lambda
$4.95
B300 SXM
Shadeform
$4.95
B300 aggregated
TensorDock
$4.99
B300 SXM
CoreWeave
$5.20
B300 HGX
Vast.ai
$5.49
B300 spot

Calculate Your Corvex GPU Costs

Enter your workload — GPU hours, model type, and training vs inference — to compare Corvex across all 34 providers in GridStackHub's network.

Open GPU Calculator → View All Provider Prices

❓Frequently Asked Questions

Is Corvex the same as CoreWeave?+
Corvex is a separate entity but shares CoreWeave's infrastructure backbone — effectively a subsidiary or sibling focused on competitive pricing. Corvex's GPU instances run on the same hardware and data centers as CoreWeave, but with different pricing tiers. If you're comparing CoreWeave vs Corvex for a given GPU, Corvex typically offers better rates on both on-demand and reserved pricing.
Why is Corvex B300 the cheapest in the network at $4.48/hr?+
Corvex positions B300 as a competitive entry point for Blackwell workloads — undercutting CoreWeave ($5.20), Lambda ($4.95), and Vast.ai ($5.49) on the same GPU. For teams running inference on 70B+ parameter models where B300's 192GB memory eliminates quantization requirements, Corvex's $4.48/hr is now the benchmark rate in GridStackHub's tracking network.
Is the H100 reserved at $1.59/hr worth the 1-year commitment?+
For teams with predictable GPU usage, H100 at $1.59/hr reserved is a strong deal — 47% below network average and below most providers' spot pricing. At 730 hours/month, that's $1,160/mo vs $2,184/mo on-demand. Over 12 months that's $13,920 savings vs on-demand. Only commit if your workload is predictable — cancellation terms matter.
Does Corvex support multi-GPU cluster configurations?+
Yes — Corvex supports multi-GPU configurations with InfiniBand interconnect for distributed training. Check corvex.io for current cluster availability and multi-GPU pricing. GridStackHub's calculator at /calculator can model multi-GPU cluster costs across providers.
How does Corvex compare to RunPod for H100?+
RunPod H100 on-demand is $2.49/hr — cheaper than Corvex's $3.15/hr for on-demand. However, Corvex offers the $1.59/hr reserved option that RunPod doesn't match. For workloads under 4 hours/day, RunPod on-demand wins. For workloads above 8 hours/day, Corvex reserved breaks even vs RunPod on-demand within 3 months.