
Takeaways by Saasverse AI
- FriendliAI | Seed Extension | $20 Million | AI Inference Optimization.
- Led by Capstone Partners, with participation from Sierra Ventures, Alumni Ventures, KDB, and KB Securities.
- Funding will drive North American and Asian expansion, R&D investment, and GPU infrastructure scaling.
FriendliAI, a pioneering company in AI inference optimization, has raised $20 million in an extended seed round led by Capstone Partners. Other participants in the round include Sierra Ventures, Alumni Ventures, KDB, and KB Securities. This latest funding will enable the company to expand its market presence in North America and Asia, bolster its research and development capabilities, and scale its GPU infrastructure to meet growing demand.
Inference represents one of the most resource-intensive stages in deploying large language models (LLMs), with IBM research indicating that it consumes 90% of operational time. FriendliAI addresses three critical challenges enterprises face in AI production environments: infrastructure costs, inference latency, and operational complexity. By leveraging its cutting-edge platform and proprietary technologies, FriendliAI promises to reduce GPU costs by up to 90% while significantly improving inference speed—all without compromising model accuracy.
Central to its innovation is the "continuous batching" technology, which optimizes how inference requests are processed. Traditional batching methods often delay individual requests until an entire batch is processed, leading to inefficiencies. FriendliAI's approach minimizes this latency, boosting throughput for LLMs by over tenfold in certain scenarios. Additionally, the company supports advanced features like N-gram speculative decoding, which reutilizes historical response data to enhance generation efficiency.
FriendliAI’s platform currently supports one-click deployment of over 430,000 models from Hugging Face and recently partnered with LG AI Research to launch EXAONE 4.0, an API for large language models. With three product offerings—Friendli Container for private GPU clusters and two cloud-based solutions for open-source AI models and custom LLMs—the company is well-positioned to cater to a variety of enterprise needs. Its client base includes 25-30 major organizations, and revenue is projected to grow by 600% this year.
“ FriendliAI's approach to inference optimization is particularly timely as the market shifts from AI experimentation to scaled production deployments. By addressing latency and cost issues, the company is poised to redefine the AI inference landscape, offering substantial value to enterprises grappling with growing computational demands. ” Saasverse Analyst comments
Saasverse Insights
the AI inference market is set for exponential growth, driven by the increasing adoption of generative AI technologies in production environments. With its technical prowess and a robust business model, FriendliAI is well-positioned to capitalize on this trend, marking itself as a key player in the evolving AI ecosystem.