FriendliAI, a burgeoning startup in the realm of artificial intelligence, has successfully raised $20 million in a seed extension round aimed at enhancing its AI inference platform. This funding round was led by Capstone Partners and saw participation from notable investors including Sierra Ventures, Alumni Ventures, KDB Investment, and KB Securities. The announcement marks a significant milestone for FriendliAI, which is poised to capitalize on the rapidly expanding market for AI inference.
AI inference represents the critical “last mile” of artificial intelligence, where users interact with AI models, such as chatbots, to receive responses generated by complex algorithms. As organizations increasingly transition from experimental phases of AI to full-scale production deployments, the demand for efficient and cost-effective AI inference solutions has surged. According to Byung-Gon Chun, CEO of FriendliAI, an estimated 80% to 90% of GPU resources are now dedicated to inference tasks, underscoring the importance of optimizing this aspect of AI operations.
Founded in 2021 by Chun, who previously led research on accelerating AI model execution at Seoul National University, FriendliAI has quickly established itself as a key player in the AI inference landscape. The company shifted its focus to AI inference in 2022, even prior to the launch of OpenAI’s ChatGPT, and has since developed a robust platform designed to streamline the deployment of AI models. In late 2023, FriendliAI relocated its headquarters to Redwood City, California, a strategic move that positions it within the heart of Silicon Valley’s tech ecosystem.
The core mission of FriendliAI is to enable companies to run AI model inference faster, cheaper, and more simply. The startup’s proprietary inference engine leverages deep algorithmic and system-level optimizations on GPUs, allowing for significant reductions in operational costs while enhancing performance. Chun asserts that FriendliAI’s technology can deliver up to 90% savings on GPU costs, making it an attractive option for businesses looking to optimize their AI infrastructure.
As the generative AI landscape continues to evolve, FriendliAI’s platform is designed to meet the growing needs of enterprises deploying large language models (LLMs) and other AI systems. The startup offers a usage-based pricing model, where customers are charged based on GPU hours consumed or the number of tokens or image steps processed. This flexible pricing structure allows organizations to scale their AI operations without incurring prohibitive costs.
FriendliAI provides three distinct deployment options tailored to the varying needs of its clients: dedicated endpoints, serverless APIs, and containers. Dedicated endpoints allocate GPUs exclusively to a client, ensuring optimal performance for high-demand applications. Serverless APIs offer access to popular AI models without the need for extensive infrastructure management, while containers allow clients to run AI models directly within their own environments.
One of the standout features of FriendliAI’s platform is its innovative “continuous batching” technique. Traditional methods of handling LLM inference traffic often struggle with the dynamic and uneven nature of request patterns. Continuous batching addresses this challenge by enabling the system to dynamically add new requests to a batch or remove completed ones at a granular level. This approach not only improves GPU utilization but also enhances overall system efficiency, particularly during periods of fluctuating demand.
FriendliAI’s technology has garnered significant traction in the market, with the company reporting rapid growth in both usage and revenue in 2025. Although specific revenue figures have not been disclosed, Chun anticipates that revenue will increase six to seven times compared to 2024, driven by the accelerating adoption of generative AI in production environments. While the company is not yet profitable, its focus remains on scaling efficiently and maintaining strong gross margins as it expands its capacity.
The startup boasts a diverse customer base that includes both startups and large enterprises. Among its notable partnerships is a collaboration with LG Electronics, where FriendliAI serves as the exclusive API provider for EXAONE, LG AI Research’s foundation model. The company currently works with approximately 25 to 30 large clients, including Scatter Lab, a social chatbot company in Korea that has reportedly achieved substantial reductions in infrastructure costs by utilizing FriendliAI’s platform to operate multiple LLMs.
Chun emphasizes that the AI inference market is experiencing explosive growth as more organizations transition from experimentation to production deployment. This shift presents a unique opportunity for FriendliAI to position itself as a leader in the space, providing essential tools and technologies that facilitate the efficient deployment of AI models.
In an increasingly competitive landscape, FriendliAI differentiates itself through its specialization in AI inference. While competitors like Fireworks AI also operate in this domain, Chun believes that FriendliAI’s proprietary inference engine and broad model coverage set it apart. The platform supports both open-source and custom models, with over 420,000 models directly deployable from Hugging Face, a leading repository for machine learning models.
Eun-gang Song, a partner at Capstone Partners, commended FriendliAI for its exceptional technical innovation in the AI inference space. He noted that the platform’s ability to deliver superior performance while reducing costs makes it an ideal partner for enterprises looking to scale their AI operations effectively.
As the demand for AI solutions continues to rise, FriendliAI is well-positioned to play a pivotal role in shaping the future of AI inference. The company’s commitment to developing cutting-edge technologies and optimizing AI deployment processes aligns with the broader trends in the industry, where efficiency and cost-effectiveness are paramount.
Looking ahead, FriendliAI’s trajectory appears promising. With a solid foundation built on innovative technology, strategic partnerships, and a clear understanding of market needs, the startup is poised to capitalize on the growing interest in generative AI and AI inference solutions. As organizations increasingly seek to harness the power of AI, FriendliAI’s offerings will likely become indispensable tools for driving successful AI initiatives.
In conclusion, FriendliAI’s recent $20 million seed extension round underscores the growing importance of AI inference in the broader AI ecosystem. As the company continues to refine its platform and expand its client base, it stands at the forefront of a rapidly evolving market, ready to meet the demands of enterprises seeking to leverage AI for competitive advantage. The future of AI inference is bright, and FriendliAI is positioned to lead the charge in delivering innovative solutions that empower organizations to unlock the full potential of artificial intelligence.