Silicon Flow, a leading third-party Model-as-a-Service (MaaS) platform, recently announced the completion of its Series B financing round exceeding RMB 2 billion (approximately $294 million USD). NIO Capital participated as a key investor in the round.

Yao Li, Partner at NIO Capital, noted that as an investment firm deeply engaged in technological innovation and committed to seizing opportunities in the AI wave, NIO Capital has always been focused on the paradigm shift in large models from “parameter competition” to “engineering deployment.” He added that the firm believes the definitive opportunity in the AI era lies in building efficient and accessible foundational infrastructure that can power real‑world applications at scale.

Li further commented: “Silicon Flow, with its outstanding proprietary inference engine and heterogeneous computing orchestration capabilities, has successfully defined and practiced the Token Factory model, taking a leading position in adaptation to domestically manufactured AI chips and the MaaS market. We are highly confident in its potential to become a globally leading token supply platform, and we look forward to leveraging our industrial ecosystem resources to help the company bring its efficient AI production capacity to more real‑world scenarios.”

That conviction is backed by strong commercial traction. Over the past year, Silicon Flow has achieved explosive growth in the enterprise segment. Through its Token Factory model delivering high‑performance MaaS, the platform now processes trillions of daily token calls and serves more than 10 million users and 10,000 enterprise customers. Annual revenue grew more than tenfold year‑on‑year, while overseas monthly revenue reached several million dollars. Its client roster spans core state‑owned enterprises in energy, finance, and transportation; public utilities and national research institutes; multiple telecom operators and large‑scale regional AI computing centers; cloud service giants, leading LLM developers, and major internet companies; as well as hundreds of top‑tier AI application developers and tool providers.

Behind this momentum is a decade‑deep pedigree in AI systems software. From its inception, Silicon Flow foresaw the need for a Token Factory model – a systematic approach to boosting token production efficiency and driving affordability through economies of scale. Today, that vision has materialized into a world‑class, fully self‑developed token production line, anchored by three core technological pillars.

First, atomic‑level engineering optimization that unleashes peak compute efficiency. Its proprietary inference engine integrates advanced techniques including PD separation, KV cache management, expert parallelism, and pipeline parallelism. It supports mainstream models such as DeepSeek, Qwen, GLM, and Kimi, and delivers high‑throughput, low‑latency commercial‑grade performance across a diverse range of chips – from NVIDIA to domestic alternatives like Ascend, Muxi, and Moore Threads.

Second, unified heterogeneous compute orchestration that tackles the supply‑demand matching challenge. The platform enables intelligent scheduling, elastic scaling, and cross‑region resource allocation, significantly boosting utilization while driving cost to the extreme. This gives enterprise customers the flexibility to handle diverse business scenarios and massive token production requirements. Silicon Flow has also forged strategic partnerships with numerous computing power providers both at home and abroad.

Third, rapid adaptation for over 160 models, offering full‑modal choice. The platform encapsulates model adaptation into reusable modules, allowing the latest open‑source models to be deployed in record time. Its model inventory far outpaces that of any other third‑party MaaS provider, covering a comprehensive range of modal tasks and giving enterprises ample options for complex application needs.

Leveraging its proprietary inference engine and orchestration system, Silicon Flow’s solution can turn any bare‑metal computing power into a standardized Token Factory with a single click. This helps compute‑asset holders overcome persistent pain points – complex heterogeneous adaptation, low utilization, and poor return on assets – and maximize the commercial value of their infrastructure.

Looking ahead, Silicon Flow noted that its Series B investor lineup represents the entire AI value chain, from chips and models to cloud platforms and application ecosystems. This full‑spectrum consortium not only provides capital but also unlocks deep synergies in scenarios, compute resources, models, and market access. The company said it will accelerate its journey toward becoming “the world’s best token supply platform” and remain committed to making AI technology universally beneficial to humanity.