DeepInfra
Inference cloud for open-source AI models
DeepInfra provides infrastructure for running AI inference on open-source models. Founded in 2022 by former imo engineers, it operates its own hardware in US data centres. The company raised $107 million in Series B funding in May 2026 to scale its services.
- Raised $107M Series B in May 2026
- Runs own Nvidia B300 GPU infrastructure
- Supports over 200 open-source AI models
- Founded by former imo messenger engineers
- Zero data retention policy for privacy
What it is
DeepInfra offers a cloud platform dedicated to AI inference, the process where models generate outputs from inputs. The service focuses on open-source models, providing developers with APIs for tasks such as text generation, image creation, and speech recognition. The company claims to own its infrastructure, including GPU hardware and data centre operations, to ensure performance and cost control. It serves both startups and enterprises with pay-as-you-go pricing and no long-term contracts.
How it came about
The team behind DeepInfra previously built backend infrastructure for imo messenger, a platform with over one billion downloads. This experience with high-scale message processing informed their approach to AI infrastructure. The company was founded in September 2022. In May 2026, it announced a $107 million Series B funding round. This capital injection aims to expand its inference cloud capabilities and support growing token volumes.
The compute
The platform runs on proprietary infrastructure optimised for inference workloads. It supports over 200 models, including recent releases like DeepSeek-V4 and Kimi K3. Hardware includes Nvidia B300 GPUs, available through a dedicated cluster product called DeepCluster. The system is designed for low latency and high throughput, handling trillions of tokens. Security features include zero data retention policies and SOC 2 and ISO 27001 certifications.
Where the GPUs are
Compute resources are hosted in secure data centres located within the United States. The company emphasises that it owns and operates this hardware rather than relying on third-party cloud providers. This setup allows for direct control over uptime, with a service level agreement targeting 99.982 per cent availability. The physical infrastructure is tailored to meet the specific demands of large language model inference.
Customers and deals
DeepInfra targets developers and companies seeking alternatives to proprietary model lock-in. Its client base ranges from early-stage startups to large enterprises requiring scalable AI solutions. The platform provides hands-on technical support and custom solutions for specific needs. Users can access models for various applications, including coding assistants, financial research tools, and creative media generation.
Money and ownership
Investors in the May 2026 Series B round include 500 Global, Georges Harik, A.Capital Ventures, Crescent Cove, Felicis, Nvidia, Peak6, Samsung Next, Supermicro, and Upper90. The round raised $107 million. The company reports a twenty-five-fold increase in token volume since its Series A. Pricing models include discounted rates for specific models and hourly rates for dedicated GPU access.
Its place in AI
DeepInfra positions itself as a key provider for the open-source AI ecosystem. By offering efficient infrastructure for models like Llama, Mistral, and DeepSeek, it supports the broader adoption of non-proprietary AI. The company highlights the importance of owning the full stack, from hardware to API, to deliver competitive performance. It aims to reduce costs for developers while maintaining high reliability.
Questions people ask about DeepInfra
What type of AI services does DeepInfra provide?
DeepInfra specialises in AI inference, allowing users to run open-source models via API. It supports text generation, image creation, speech recognition, and other tasks without requiring users to manage their own hardware.
Who are the main investors in DeepInfra?
The company's Series B investors include 500 Global, Georges Harik, Nvidia, Samsung Next, and Supermicro. Other participants include A.Capital Ventures, Felicis, and Upper90, among others.
Where is DeepInfra's infrastructure located?
DeepInfra operates its own hardware in secure data centres based in the United States. The company owns this infrastructure to ensure better control over performance, cost, and reliability compared to public cloud options.
Does DeepInfra store user data?
No, DeepInfra has a zero retention policy. Inputs, outputs, and user data are not stored after processing. The platform is also SOC 2 and ISO 27001 certified for security and privacy compliance.
Good reading and sources
There is no English Wikipedia article on DeepInfra yet, so the facts on this page come from the sources listed above: deepinfra.com/about, deepinfra.com/. 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.