News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory鈥檚 NVIDIA & AMD Enterprise ExpansionDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory鈥檚 NVIDIA & AMD Enterprise Expansion
EveryNeocloud
Vast.ai

Vast.ai

Marketplace for distributed GPU cloud compute

Vast.ai operates a two-sided marketplace connecting users with underutilized GPU hardware from independent hosts and data centres. Founded in 2016 by Jake Cannell and Christian Horne, it offers cheaper access to AI compute than major hyperscalers, supporting training and inference workloads.

Share on WhatsApp

  • Founded in 2016 by Jake Cannell and Christian Horne
  • Offers Nvidia GB200 Superchips for rent
  • Over 17,000 GPUs available on the platform
  • Headquarters in Los Angeles and San Francisco
  • SOC 2 Type I certified as of 2025

What it is

Vast.ai is a platform that allows users to rent GPU computing power from a distributed network of hosts. These hosts range from individuals with idle hardware to professional data centres. The company acts as an intermediary, providing the software layer to manage instances, handle payments, and ensure security. It aims to lower the cost of AI development by offering rates significantly below those of major cloud providers. Users can access bare-metal machines, virtual machines, or serverless inference endpoints. The service supports various AI tasks, including model training, fine-tuning, and image generation.

How it came about

The concept originated from essays by Jake Cannell on LessWrong, arguing that intelligence scales with compute. He collaborated with Christian Horne, who shared the view that centralized control of compute would limit AI progress. They incorporated Vast.ai in Delaware in June 2016. For two years, they built the platform internally before launching quietly to early adopters. Growth accelerated as independent hosts joined, validating the model of monetizing idle hardware. Travis Cannell joined as COO in April 2022 to support scaling. The company opened its first office in Los Angeles in July 2024 and a second in San Francisco in 2025.

The compute

The platform supports a wide range of Nvidia GPUs, including the recent GB200 Superchip. It manages heterogeneous hardware through Docker-based instance management, allowing it to operate across different network conditions. Users can provision secure cloud instances from certified data centre partners or choose cheaper options from individual hosts. The system includes a search interface and pricing engine to help users find suitable hardware. It offers tools for AI agents to autonomously procure compute. Software support includes CLI, Python SDK, and API docs for integration.

Where the GPUs are

Compute resources are distributed globally, sourced from independent hosts and professional data centres. The company itself is based in the United States, with headquarters in Los Angeles and an engineering hub in San Francisco. It does not own the hardware but certifies certain data centre partners for its Secure Cloud offering. This hybrid approach allows it to scale without heavy capital expenditure on infrastructure. The distributed nature means availability varies by region and host.

Customers and deals

The platform serves developers, researchers, and enterprises who need affordable GPU access. It targets those priced out by hyperscalers, including graduate students and small startups. Use cases include AI agent development, fine-tuning models, and batch data processing. The company highlights that its model helps democratize access to AI research, allowing independent labs to compete. It also supports enterprise customers through its Secure Cloud product, which meets higher compliance standards.

Money and ownership

Vast.ai is a private startup. It achieved SOC 2 Type I certification in 2025. The company reported 310 per cent growth in 2023, with over 17,000 GPUs on the platform. It employs more than 40 people. Revenue comes from transaction fees on compute rentals. There is no public information on venture funding rounds or valuation in the provided material. The business model relies on volume and the spread between host earnings and user payments.

Its place in AI

Vast.ai positions itself as a counterweight to centralized cloud providers. Its founders argue that distributed compute prevents a few large companies from controlling the future of AI. By lowering barriers to entry, it enables a wider diversity of researchers to contribute to AI safety and alignment. The platform supports the agentic economy, where AI systems can buy their own compute. It plays a role in keeping the AI ecosystem open and competitive.

Questions and criticism

Security and reliability can vary since much of the supply comes from individual hosts. The company addresses this with its Secure Cloud tier for sensitive workloads. The reliance on third-party hardware means quality control is managed through user ratings and platform verification rather than direct ownership.

Questions people ask about Vast.ai

What kind of hardware can I rent on Vast.ai?

You can rent a variety of Nvidia GPUs, including the latest GB200 Superchips. The hardware comes from both individual hosts and professional data centres, offering a mix of price points and reliability levels.

How does Vast.ai differ from AWS or Azure?

Vast.ai is a marketplace that aggregates unused GPU capacity from many sources, often resulting in lower prices. Unlike hyperscalers, it does not own all the hardware, allowing for a more distributed and cost-effective model for developers.

Is my data safe on Vast.ai?

For sensitive workloads, Vast.ai offers Secure Cloud with certified data centre partners. Standard instances from individual hosts may have varying security levels, so users should choose based on their data sensitivity needs.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Vast.ai yet, so the facts on this page come from the sources listed above: vast.ai/about. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryNeocloud wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.