Browse GPUs

On-Demand GPU Compute, Settled on Solana

Calor lets you rent NVIDIA H100, A100 & RTX 4090 GPUs in one click — pay in SOL, earn Calor, and run workloads at 85% below cloud market rate.

pattern
pattern

Trusted by ML engineers, render studios and Solana-native teams running real workloads.

How it works

From Wallet to Running GPU in Minutes

Connect Your Wallet

Link Phantom, Solflare or Backpack. Your rentals and Calor rewards are tied to your Solana wallet.

Pick a GPU

Choose NVIDIA H100, A100, RTX 4090, 4080 & 3090 across 5 regions. Live pricing, hourly billing.

Pay in SOL

Settle instantly on Solana (devnet in beta). Pay in SOL or redeem Calor for a protocol-subsidized rate.

Provision in Seconds

Your instance spins up automatically. Add your SSH key and connect — no manual setup, no queues.

SSH or Web Terminal

Connect over SSH or the in-browser terminal. Full root access for training, inference or rendering.

Earn Calor

Every GPU job burns Calor and rewards stakers. Stake for priority routing to high-demand clusters.

Connect Your Wallet

Link Phantom, Solflare or Backpack. Your rentals and Calor rewards are tied to your Solana wallet.

Provision in Seconds

Your instance spins up automatically. Add your SSH key and connect — no manual setup, no queues.

Pick a GPU

Choose NVIDIA H100, A100, RTX 4090, 4080 & 3090 across 5 regions. Live pricing, hourly billing.

SSH or Web Terminal

Connect over SSH or the in-browser terminal. Full root access for training, inference or rendering.

Pay in SOL

Settle instantly on Solana (devnet in beta). Pay in SOL or redeem Calor for a protocol-subsidized rate.

Earn Calor

Every GPU job burns Calor and rewards stakers. Stake for priority routing to high-demand clusters.

Use Cases

What People Run on Calor

From LLM training to real-time inference and rendering — on bare-metal NVIDIA GPUs.

Train Faster on H100 Clusters

LLM Training & Fine-Tuning

Spin up multi-GPU H100 nodes with 80GB HBM3 and NVLink. Train and fine-tune large models at 3,958 TFLOPS FP16 — billed by the hour.

Train Faster on H100 Clusters

LLM Training & Fine-Tuning

Spin up multi-GPU H100 nodes with 80GB HBM3 and NVLink. Train and fine-tune large models at 3,958 TFLOPS FP16 — billed by the hour.

Serve Models at Low Latency

Real-Time Inference

Deploy inference endpoints on A100 40/80GB across 5 regions with ~400ms Solana settlement. Scale up and down, billed by the hour.

Serve Models at Low Latency

Real-Time Inference

Deploy inference endpoints on A100 40/80GB across 5 regions with ~400ms Solana settlement. Scale up and down, billed by the hour.

Render & Simulate at Scale

Rendering & Batch Compute

Burst to RTX 4090 and 3090 fleets for rendering, simulation and batch compute. Pay in SOL — no commitments, no idle costs.

Render & Simulate at Scale

Rendering & Batch Compute

Burst to RTX 4090 and 3090 fleets for rendering, simulation and batch compute. Pay in SOL — no commitments, no idle costs.

Why Calor

Why Choose Calor

The GPU network built on Solana for price, transparency and instant settlement.

Cloud providers

Time to first running GPU

Time to first running GPU

Minutes

Minutes

Hours to days

Hours to days

Billing transparency

Billing transparency

Hourly, fully on-chain

Hourly, fully on-chain

Opaque invoices, net-30 terms

Opaque invoices, net-30 terms

Payment method

Payment method

On-chain settlement

On-chain settlement

Deflationary rewards

Deflationary rewards

GPU availability

GPU availability

H100, A100, RTX 4090 — 5 regions

H100, A100, RTX 4090 — 5 regions

Limited stock, long queues, region locks

Limited stock, long queues, region locks

Scaling & cost

Scaling & cost

Pay only for hours used

Pay only for hours used

Idle costs and reserved-instance lock-in

Idle costs and reserved-instance lock-in

Protocol-Subsidized Rates

Redeem Calor for H100 and A100 time at up to 85% below market rate.

Burn-on-Use Deflation

Every GPU job triggers an SPL burn. Supply shrinks as the network does real work.

Staking Priority

Stake Calor for priority routing to high-demand NVIDIA clusters.

From the network

Teams Already Running on Calor


From training runs to render farms, see how teams put Calor GPUs to work.

"We moved our fine-tuning jobs to Calor H100 nodes and cut cloud spend by 80%. Provisioning that took days now takes minutes."

Sarah Chen

ML Lead, NovaFlow

"On-chain hourly billing means we only pay for the GPU hours we use. No opaque invoices, no net-30 terms."

Priya Sharma

Infra Engineer, Vertex AI

"Our render farm bursts onto RTX 4090 fleets whenever we need it. Paying in SOL settles in seconds — no invoices, no net-30."

Marcus Okoro

Studio Lead, Lumora

"Staking Calor gets us priority routing to H100 clusters, and burn-on-use keeps supply tight. Calor is a game changer."

David Nkosi

Founder, ScaleForge

Your Questions

Frequently Asked Questions

Everything you need to know about renting GPUs and using Calor on Calor.

What can I run on an Calor GPU?

How do I pay — in SOL or Calor?

How do I connect to my rented GPU?

Which GPUs and regions are available?

Is my data and SSH key secure?

The Future of Compute is On-Chain

On-demand NVIDIA H100, A100 & RTX GPUs — provisioned in seconds, billed by the hour, settled on Solana.

pattern

The Future of Compute is On-Chain

On-demand NVIDIA H100, A100 & RTX GPUs — provisioned in seconds, billed by the hour, settled on Solana.

pattern

The Future of Compute is On-Chain

On-demand NVIDIA H100, A100 & RTX GPUs — provisioned in seconds, billed by the hour, settled on Solana.

pattern

Create a free website with Framer, the website builder loved by startups, designers and agencies.