Key Takeaways
- • DeepInfra closed a $107 million Series B funding round co-led by 500 Global and Georges Harik (early Google engineer), with participation from NVIDIA, Samsung Next, Supermicro, Felicis, Peak6, Crescent Cove, A.Capital Ventures, and Upper90.
- • The platform processes nearly five trillion tokens per week and is purpose-built for high-throughput AI inference at production scale, positioning the company at an inflection point as enterprise workloads shift from model training to inference deployment.
- • DeepInfra was founded by the team behind imo, a messenger application that scaled to over 200 million users globally, bringing significant operational expertise to infrastructure scaling.
- • The funding will accelerate global capacity expansion and support the rising demand for open-source and agent-driven AI workloads, which require continuous, high-volume inference capabilities that traditional cloud platforms were not architected to deliver.
- • CEO Nikola Borisov emphasized that open-source models are achieving parity with proprietary systems while reducing costs, and agent-based systems are driving persistent demand for inference infrastructure—positioning inference as the new system constraint rather than a marginal layer.
Summary
DeepInfra, a purpose-built inference cloud platform, announced $107 million in Series B funding on May 4, 2026, reflecting the market's recognition of inference as a critical bottleneck in production-scale AI deployment. The round was co-led by 500 Global and Georges Harik, one of Google's earliest engineers, with substantial participation from both strategic and financial investors including NVIDIA, Samsung Next, Supermicro, Peak6, Felicis, Crescent Cove, A.Capital Ventures, and Upper90.
The company currently processes nearly five trillion tokens per week and enables enterprises and scaleups to operate open-source and agent-driven AI workloads with improved cost efficiency, performance, and security posture. DeepInfra was built by the founding team behind imo, a messenger application that reached 200 million global users, bringing proven infrastructure scaling expertise to the AI stack.
According to Nikola Borisov, co-founder and CEO, the funding cycle marks an inflection point in enterprise AI adoption. "When we launched nearly four years ago, we believed inference would become the dominant driver of enterprise AI workloads – and we are now at this inflection point," Borisov stated. He highlighted that open-source models are rapidly achieving cost and performance parity with proprietary systems, while agent-based architectures are generating continuous, high-volume inference demand. This shift fundamentally changes the infrastructure problem: inference has evolved from a thin computational layer into the system constraint that will define the majority of production workloads.
The investment underscores 500 Global's broader thesis across the AI infrastructure stack, as traditional cloud platforms were not architected for always-on, distributed inference at scale. DeepInfra's Series B capital will fund global capacity expansion to serve this emerging demand while maintaining the cost, performance, and security advantages required for enterprise adoption of open-source and agent-driven AI systems.
The company currently processes nearly five trillion tokens per week and enables enterprises and scaleups to operate open-source and agent-driven AI workloads with improved cost efficiency, performance, and security posture. DeepInfra was built by the founding team behind imo, a messenger application that reached 200 million global users, bringing proven infrastructure scaling expertise to the AI stack.
According to Nikola Borisov, co-founder and CEO, the funding cycle marks an inflection point in enterprise AI adoption. "When we launched nearly four years ago, we believed inference would become the dominant driver of enterprise AI workloads – and we are now at this inflection point," Borisov stated. He highlighted that open-source models are rapidly achieving cost and performance parity with proprietary systems, while agent-based architectures are generating continuous, high-volume inference demand. This shift fundamentally changes the infrastructure problem: inference has evolved from a thin computational layer into the system constraint that will define the majority of production workloads.
The investment underscores 500 Global's broader thesis across the AI infrastructure stack, as traditional cloud platforms were not architected for always-on, distributed inference at scale. DeepInfra's Series B capital will fund global capacity expansion to serve this emerging demand while maintaining the cost, performance, and security advantages required for enterprise adoption of open-source and agent-driven AI systems.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.