Salad Technologies, Inc.
Salad builds an orchestration cloud for AI inference and other compute workloads by aggregating idle consumer and enterprise GPUs and scheduling containerized workloads across distributed infrastructure. Its stated aim is to make cloud compute more affordable, plentiful, and people-powered than conventional data-center clouds.
At a Glance
- AI startups
- Enterprise software and Fortune 500 organizations
- Developers and engineering teams
- Researchers and scientific-computing users
- +4 more
AI Tools by Salad Technologies, Inc.
(1)
Salad
Distributed GPU Cloud for Containers
Discussions
No discussions yet
Be the first to start a discussion about Salad Technologies, Inc.
Latest News
Salad.com and Golem Network partner to test integrating Web2 workloads with decentralized infrastructure
RNP-023 approved: Salad joins Render Network as an exclusive subnet
Render integration milestones go live: customers can pay SaladCloud with RENDER and Chefs can receive RENDER rewards
Earning-rate changes announced for September 12, including higher RTX 5090/4090 rates and increased CPU payouts
Products & Services
OpenAI-compatible endpoint for open-weight models at ai.salad.cloud/v1, with per-token/per-request billing or a flat monthly plan.
Managed orchestration for Docker GPU workloads: users specify an image, hardware requirements, and replica count, while Salad schedules, monitors, and reallocates workloads.
On-demand, interruptible compute drawn from idle consumer GPUs worldwide, with four priority tiers and per-second billing.
Reserved bare-metal servers and virtual machines on contract, including RTX PRO 6000 Blackwell capacity, for predictable capacity, large VRAM, and data-center placement.
Market Position
Salad positions itself as a lower-cost alternative to centralized hyperscalers and conventional GPU clouds by using a globally distributed, crowdsourced supply of idle consumer GPUs and automated reallocation. It competes with hyperscalers such as AWS, Microsoft Azure, and Google Cloud and with GPU-cloud/compute marketplaces such as Vast.ai, CoreWeave, and decentralized providers including Golem and Render; its differentiation is the scale of its consumer-GPU network, low per-GPU pricing, container orchestration, and the ability to extend the same stack to dedicated and private infrastructure.
Leadership
Founders
Bob Miles
Founder and CEO. Salad's 2020 funding announcement describes him as an entrepreneur who previously worked as an engineer and product manager in the technology sector; Origin Ventures says it first met him building Salad in Utah in 2018.
Executive Team
Bob Miles
Founder and CEO
Technology entrepreneur; previously an engineer and product manager in the technology sector.
Kyle Dodson
Chief Technology Officer
Salad CTO; quoted in the Golem partnership announcement on Salad's compute-orchestration architecture and crypto-payment engineering work.
Board of Directors
Founding Story
Salad began as a crypto/compute-sharing experiment intended to monetize the large amount of idle processing capacity in consumer gaming PCs. The initial vision was to let people voluntarily share spare compute, storage, and bandwidth in exchange for meaningful rewards, creating a distributed alternative to centralized cloud providers; the company later evolved into an orchestration cloud for business workloads.
Business Model
Revenue Model
Two-sided infrastructure marketplace: businesses pay for running container replicas by the second on Community Cloud, while Salad pays GPU providers for idle capacity and retains the platform/service margin. Salad also sells pay-per-use APIs, contract-priced reserved Dedicated infrastructure, and per-deployment licenses for Private Cloud; it offers volume discounts and committed-spend agreements.
Pricing Tiers
Highest protection from preemption; still subject to node disconnection and automatic reallocation.
May be preempted by higher-priority groups.
Lower-cost interruptible tier.
Lowest-cost, most interruptible tier; intended for retryable batch and stateless inference behind a queue.
General-purpose, CPU-optimized, and memory-optimized configurations.
Reserved hardware or VMs.
Runs on customer-owned infrastructure.
Target Markets
- AI startups
- Enterprise software and Fortune 500 organizations
- Developers and engineering teams
- Researchers and scientific-computing users
- Model and API providers
- Game, rendering, and media companies
- AI model inference and serving
- Batch AI transcription and call-center analytics
- Image and video generation
- LLM and agent workloads
- Model training and fine-tuning
- Computer vision
- Civitai
- Blend
- Klyne.ai
- Undetectable AI