SaladCloud pricing: GPU rental rates
Visit SaladCloud →Updated July 2026 · day 14 of the ledger · verified 2026-07-18
The cheapest GPU on SaladCloud right now is the RTX 3090 at $0.12/hr (community cloud, verified 2026-07-18). SaladCloud runs container workloads on a distributed network of consumer PCs whose owners share their GPUs for rewards. That model makes it one of the cheapest sources of consumer cards like the RTX 5090, 5080, 4090 and 3090, with the tradeoff that capacity comes from individual machines rather than datacenters. The public pricing table lists batch-priority rates that bundle a baseline container of 8GB RAM and 4 vCPUs; we track those rates daily and label them community tier.
Live SaladCloud prices (2026-07-18)
| GPU | Tier | $/hr | Verified |
|---|---|---|---|
| RTX 3090 | community cloud | $0.12 | 2026-07-18 |
| RTX 3090 Ti | community cloud | $0.15 | 2026-07-18 |
| RTX 4090 | community cloud | $0.20 | 2026-07-18 |
| RTX 5080 | community cloud | $0.22 | 2026-07-18 |
| RTX 5090 | community cloud | $0.29 | 2026-07-18 |
Cheapest listed per-GPU-hour rates verified at the source. Before storage, egress and networking.
Pricing model
Salad prices GPU container instances per hour, with the cost determined by how long each container instance runs. The published GPU table is for the batch priority level, the lowest of Salad's priority tiers, and each row includes a baseline of 8GB RAM and 4 vCPUs; instances are customizable to other GPU, vCPU and RAM shapes. Higher priority levels (lowest, low, medium, high) are priced through the site's calculator rather than a static table, so we publish the batch rates the page itself commits to.
Capacity model
SaladCloud is a distributed network: compute comes from consumer PCs running the Salad agent, not from a datacenter fleet the company owns. Workloads run as containers via the Salad Container Engine. Because nodes are consumer machines, treat capacity as interruption-tolerant and design for checkpointing; the priority-tier system exists precisely because workloads compete for node time. We label Salad offers community tier, the same label we use for other peer-hardware marketplaces, and never as firm on-demand capacity.
GPU lineup
The public table lists consumer GeForce cards only: RTX 5090, RTX 5080, RTX 4090, RTX 3090 Ti, RTX 3090, RTX 4080, RTX 4070 Ti, RTX 3080 Ti, RTX 3080, RTX 3070 Ti, RTX 3070 and RTX 3060. No datacenter GPUs (H100 class) appear on the current pricing page. CPU-only general-purpose, CPU-optimized and memory-optimized instance shapes are also published.
Who should pick SaladCloud
Pick SaladCloud for interruption-tolerant, throughput-oriented GPU work on consumer cards: batch inference, image generation, transcription and similar container jobs where the lowest cost per GPU-hour beats guaranteed node stability. It is a poor fit for workloads needing datacenter GPUs, large VRAM pools, multi-GPU nodes or strict uptime guarantees. A formal public uptime SLA for GPU workloads was not found in the pages reviewed. Not publicly documented.
FAQ
Capacity comes from consumer PCs in a distributed network, and the public pricing is tiered by priority level, so workloads should be designed as interruption-tolerant. We label Salad rates community tier, not firm on-demand.
The published batch-priority rates include the GPU plus a baseline container of 8GB RAM and 4 vCPUs. Instances are customizable to other shapes through the pricing calculator.
The current public pricing table lists consumer GeForce cards only. No H100-class datacenter rows appear on the page we verify.