DeepInfra Secures New Series B Funding
DeepInfra has raised $107 million in new Series B funding. The company is based in Palo Alto. The round was led by 500 Global and angel investor Georges Harik. It included participation from several technology and investment groups. These include A.Capital Ventures, Crescent Cove, Felicis, Nvidia, Peak6, Samsung Next, Supermicro, and Upper90.
The company says the new capital marks an important moment. It reflects a shift in how organisations use AI. It also highlights the rise of inference as the main driver of AI workloads. DeepInfra says its platform processes far more tokens than before. It has seen token processing grow by 25 times since its Series A.
The company was founded four years ago. Its goal was to focus on inference rather than training. It believed inference would dominate enterprise workloads. That belief has now become reality. Demand for low-latency and high-volume inference continues to rise.
DeepInfra says most cloud platforms were not built for continuous inference. Many were designed for mixed workloads. These workloads are often unpredictable. They do not match the constant flow needed for high-volume token generation. DeepInfra built its system to solve this gap.
High-Throughput Inference Built From The Ground Up
The company runs a cloud platform designed for AI inference. It supports open-source and proprietary models. It focuses on high-throughput performance. It aims to deliver predictable latency and lower cost. Many AI systems require fast output generation. They also need stable infrastructure for sensitive tasks.
Inference involves turning input into output. This simple idea hides massive computation needs. It takes over 100 billion operations to create a single AI token. DeepInfra supports these workloads using its own infrastructure. It owns and operates GPU clusters across eight data centres in the United States. More sites are planned worldwide.
Owning the hardware helps the company maintain stability. It avoids reliance on spot capacity. It also reduces delays created by rented systems. DeepInfra says this gives it stronger efficiency. It also gives developers more predictable performance during peak demand.
The company supports more than 190 open-source models. It offers APIs that follow OpenAI’s API format. It also includes enterprise-grade security. This includes zero data retention. It also includes SOC 2 and ISO 27001 certifications.
DeepInfra supports the rise of agentic models. These models require continuous token output. One agent task can involve more than 50 model calls. Some involve more than 100. These tasks run constantly. They also require consistent compute power. DeepInfra says its system is engineered for this demand.
Engineering Approach Anchored In Vertical Integration
DeepInfra’s vertical integration approach, showcasing unified hardware, networking, and software layers powering efficient AI inference systems. Source: Created by Ventureburn.
DeepInfra says its efficiency comes from full-stack control. It designs its systems across hardware, networking, and software. This gives it more control over throughput. It also avoids performance losses seen in general-purpose cloud systems.
The company has long experience in distributed systems. Members of the founding team built large-scale messaging systems in earlier roles. These systems supported more than 200 million users. That experience shaped its approach to inference infrastructure.
The company also collaborates with hardware and software partners. This includes work within the open AI ecosystem. It supports model frameworks, inference systems, and new distributed technologies. These tools help improve cost efficiency. They also increase the speed of model deployment.
DeepInfra says early access to next-generation GPUs provides major gains. New hardware helps deliver faster results. It also reduces cost per token. This matters for continuous workloads. It helps developers scale AI systems safely. It also helps enterprises manage large production tasks.
More News: Barocal Raises $10M To Scale Solid-State Cooling Technology
New Funding Supports Global Scaling
The new round will support three priorities. The first is expanding global compute capacity. The company plans to add new data centre locations. These sites will increase available GPU power. They will also improve reliability for worldwide users.
The second priority is improving developer tools. Developers need faster ways to build AI products. They also need stable tools for high-volume tasks. DeepInfra says it will expand support for new workflows. It will also release tools that make production workloads easier to manage.
The third priority is supporting the next wave of AI models. Open-source systems continue to advance. Agentic models require more compute power. These models also need stable systems for constant output. DeepInfra plans to support these new systems as they emerge.
The company says high-quality inference will define enterprise AI. Training still matters. However, production workloads depend on inference. They depend on reliable systems that run continuously. DeepInfra says it is building the infrastructure for that reality.
Demand for inference continues to rise. More businesses use AI in daily workflows. More products require constant token generation. More systems depend on fast responses. DeepInfra believes it is positioned to support this shift.
The company expects ongoing growth. It says the next phase will focus on expanding capacity. It also plans to strengthen its developer ecosystem. It aims to make production-grade inference accessible to organisations of all sizes.
To stay updated on crypto venture capital funding and market trends, visit our venture capital news section for more insights.
