Deepinfra Inc. raises $107M in Series B

Deepinfra Inc., a startup specializing in dedicated inference cloud infrastructure, has secured $107 million in a Series B funding round. The funds, led by 500 Global and former Google engineer Georges Harik, aim to bolster Deepinfra's global expansion efforts. Notable investors include Nvidia Corp., Samsung's venture branch Samsung Next, Supermicro Computer Inc., as well as A.Capital Ventures, Crescent Cove, Felicis, Peak6, and Upper90.

Deepinfra is redefining cloud infrastructure to better serve artificial intelligence (AI) workloads as the industry pivots from simple chatbots to sophisticated agentic workflows that operate autonomously. Co-founder and CEO Nikola Borisov noted the increasing importance of inference processes in shaping enterprise AI operations, stating that they are becoming the main constraint on AI systems as opposed to merely serving as a secondary service. With its own hardware spread across eight U.S. data centers, Deepinfra aims to improve cost efficiency and performance, handling the entire infrastructure stack for AI model deployment.

The venture focuses on addressing inefficiencies inherent in existing cloud platforms, which struggle with the continuous demands of autonomous AI agents. By employing Nvidia's Dynamo platform and advanced GPUs, Deepinfra reports generating 20 times greater inference cost efficiency compared to traditional cloud services. This efficiency is crucial as around 30% of Deepinfra’s platform usage stems from resource-intensive autonomous agents rather than conventional chatbot applications.

In the cloud infrastructure market, Deepinfra's approach contrasts with general-purpose providers by emphasizing predictable latency and reduced costs for AI deployments. By supporting over 190 open-source AI models and ensuring zero-data retention, Deepinfra distinguishes itself with a security focus conducive to enterprises with sensitive data. Nvidia’s Nemotron models are among those facilitated by Deepinfra's infrastructure, aligning with a larger trend toward open-source technology that CEO Borisov describes as fostering rapid innovation at reduced costs.

The investment reflects growing recognition of infrastructure tailored specifically for AI inference as a critical component of the next phase in AI development. According to Tony Wang of 500 Global, demand for more robust AI inference support is skyrocketing, and Deepinfra's proven expertise in global-scale systems positions it to capitalize on this trend. Future steps likely include scaling operations to meet this burgeoning demand, although specific next milestones have not been disclosed.