Salad

Salad is a distributed GPU cloud for AI and other GPU-heavy workloads, with usage-based pricing and no contracts or pre-payment. It helps teams deploy inference, transcription, image generation, and batch processing on consumer GPUs.

Other

Visit Website

Overview

Salad is a distributed GPU cloud for running AI and other compute-heavy workloads on consumer GPUs. The site positions it as a lower-cost alternative to hyperscale clouds, with pay-as-you-go pricing and no contracts or pre-payment.

The product is presented as a managed container service rather than a VM management platform. The source pages emphasize scaling inference, batch jobs, computer vision, language models, voice AI, and image generation without managing individual machines, while exposing customizable GPU, RAM, and vCPU options.

Core capabilities

Distributed GPU infrastructure

Deploy AI and other containerized workloads on a distributed cloud of consumer GPUs, with the pages highlighting inference at scale, batch processing, rendering queues, and LLM workloads.

Managed container deployment

Scale workloads without managing individual VMs or instances, using a fully managed container service and usage-based execution model.

Customizable instance sizing

Choose from customizable GPU, memory, and vCPU configurations, with pricing pages showing several GPU classes and general-purpose, CPU-optimized, and memory-optimized instance types.

Image generation workflows

Support for image generation workflows is highlighted with pre-built containers and benchmarks for Stable Diffusion XL, Flux.1-Schnell, and other image models.

Voice and transcription workloads

Voice AI pages highlight transcription, translation, captioning, and text-to-speech workloads, including the Salad Transcription API powered by Whisper Large v3.

Global node distribution

The platform is positioned as globally distributed, with geo-distributed nodes across many countries and low-latency edge placement for workloads.

Common use cases

Image generation inference
Run text-to-image or image-to-image inference on consumer GPUs when image output volume and per-image cost matter. The source highlights Stable Diffusion XL, Flux.1-Schnell, and pre-built container workflows.
Voice AI and transcription
Deploy speech-to-text, text-to-speech, translation, captioning, or subtitle workflows on GPUs instead of managed voice APIs. The source pages emphasize lower cost and the Salad Transcription API.
LLM deployment
Serve language-model workloads, including self-hosted LLMs and token-based inference, without managing your own VM fleet.
Batch processing and rendering
Run distributed batch jobs, HPC-style workloads, or rendering queues across many GPU nodes when the task can be parallelized.
Elastic production scaling
Build production AI services that need elasticity, such as teams scaling inference on demand without over-provisioning their own hardware.

Pros and Cons

Pros

Published pricing starts at very low hourly rates and is presented as usage-based with no contracts or pre-pay.
The platform covers several GPU-heavy workloads, including image generation, voice AI, transcription, computer vision, and LLM deployment.
Instance options are customizable across GPU class, RAM, and vCPU requirements.
The site describes a large distributed network with geo-distributed nodes and managed container deployment.
The transcription offering includes a dedicated API with a stated free trial for new organizations.

Cons

The source is lighter on integration details, so buyers will need to confirm supported tooling and deployment paths from the docs.
Some of the published benchmarks and lowest-price claims are workload-specific, so they should not be treated as universal across every model or instance type.

FAQ

What is Salad Cloud used for?

Salad Cloud is a distributed GPU cloud for running AI and other GPU-heavy workloads on consumer GPUs. The source pages show use cases such as image generation inference, voice AI, transcription, language models, computer vision, batch processing, and rendering queues.

How does Salad Cloud pricing work?

The pricing page shows pay-as-you-go instance pricing with no contracts or pre-pay, and the FAQ says you only pay for the time the hardware is available to your application. Charges do not start during cold boot or initialization.

Can I bring my own workloads to Salad Cloud?

The source pages indicate SaladCloud is a fully managed container service, and Salad Container Engine workloads can run alongside hybrid or multi-cloud configurations. The pages also note that additional options are available through the SaladCloud API.

Does Salad Cloud mention security or compliance?

SaladCloud is described as SOC2 certified on the home page. The same page also says its patented approach isolates customer environments and data across the network.

Is there a free trial?

The source mentions a free trial for the Salad Transcription API: a 5-audio-hour trial is automatically applied when you create your organization. The pricing page says Salad Container Engine does not have a free trial.