ⓘ This article is third-party content and does not represent the views of this site. We make no guarantees regarding its accuracy or completeness.
Recently, Approaching.AI, an AI Token production service provider, announced the completion of its Series Pre-A funding round, raising hundreds of millions of RMB.
Following the funding round, Approaching.AI will continue to increase investment in its high-efficiency AI Token production service platform (Approaching AI Token as a Service, ATaaS), with a focus on expanding computing power reserves and advancing underlying inference system development. The company will persistently deliver model output capabilities characterized by low latency, high throughput, stable structured output, reliable function calling, and predictable service quality, further enhancing its ability to supply high-quality Tokens at scale for enterprise production environments.
Approaching.AI: A Core Supplier of High-Quality Tokens in China, with ATaaS Approaching One Trillion Daily Average Token Calls
As large language model (LLM) applications enter enterprise production environments, the evaluation criteria for AI inference infrastructure are undergoing a transformation. Enterprises are no longer solely focused on computing power scale, the number of models, or interface richness; instead, they place greater emphasis on whether each API call can stably, efficiently, and predictably fulfill business requirements. At this stage, the core competitiveness of inference services is shifting from "providing models" to "producing high-quality Tokens." Metrics such as time-to-first-token (TTFT), tokens per second (TPS) output, structured output stability, function calling reliability, and service quality predictability under high-concurrency scenarios are becoming key indicators for enterprises when selecting AI infrastructure.
Approaching.AI believes that Tokens are no longer merely the basic units of LLM input and output but have become critical production factors linking model capabilities, system performance, service stability, and cost efficiency. Based on this insight, the company has proposed the industry concept of "Token as a Service" (TaaS) and developed the high-performance AI Token production service platform, ATaaS. Unlike traditional MaaS (Model as a Service), which focuses on model invocation and management, ATaaS is more focused on delivering inference efficiency for enterprise production scenarios, helping businesses acquire scalable, operable high-quality Token production capabilities.
In terms of model strategy, Approaching.AI adheres to a "fewer models, deeper optimization" approach. Rather than indiscriminately supporting hundreds of models, the company focuses on a select few high-productivity models, continuously optimizing output quality, inference efficiency, TTFT stability, and TPS performance based on real-world enterprise scenarios. For enterprise customers, the sheer number of models does not directly translate to productivity; what truly matters is whether each call can reliably support business outcomes.
Regarding system capabilities, Approaching.AI leverages heterogeneous computing power scheduling, cross-cluster cache sharing, inference link isolation, elastic scaling, and quality monitoring to transform underlying computing resources into sustainable, high-quality AI Token production capacity. Leveraging full-stack systems engineering capabilities, the company delivers more stable TTFT, high-speed output at 30–50 TPS, and reliable service guarantees while maintaining cost control.
Currently, Approaching.AI serves multiple enterprise-level customers through its ATaaS platform, which processes nearly one trillion Tokens daily. Through long-term validation in highly complex, high-concurrency business scenarios, the company has developed core capabilities for large-scale inference delivery.
A Unique and Irreplicable Talent Matrix, Supported by Both Commercial Traction and a Strong Technical Foundation to Drive ATaaS Strategy
TaaS is not an ordinary application-layer product but rather a systemic capability spanning the entire AI inference chain. It requires a team that understands enterprise customer needs, industry resources, capital pathways, and commercialization pacing, as well as deep, long-term expertise in foundational areas such as computing, storage, scheduling, caching, and inference systems. Approaching.AI's core team combines commercial execution capabilities with deep technical expertise, laying the groundwork for ATaaS to evolve from a cutting-edge concept to large-scale deployment with leading customers within just two years of the company's founding.
On the commercial and operational front, Approaching.AI has developed the organizational capability to advance both technology productization and commercial capitalization in a coordinated manner. Founder and CEO Dr. Ai Zhiyuan, a Tsinghua computer science Ph.D., brings both systems research experience and large-scale commercial expertise, having pioneered the TaaS industry logic and driven ATaaS from a technical platform to an enterprise-grade production service. President Dr. Wu Wenjie, who holds a Ph.D. in finance and a CFA charter, has senior executive experience at leading industry and capital institutions, having led numerous benchmark investments and acquisitions. She oversees the company's strategy, internal controls, and global operations. Chairman Ren Xuyang, an early entrepreneurial veteran, has founded multiple companies and supports the organization through industry insight, organizational development, capital coordination, and ecosystem resource integration.
On the technical and research front, Approaching.AI is backed by more than two decades of accumulated expertise from Tsinghua University's Institute of High Performance Computing (IHPC). The company has completed a process whereby related Tsinghua technologies have been valued and contributed as capital increase through a technology transfer. These technologies, developed over the long term by the research teams of Academician Zheng Weimin, Professor Wu Yongwei, Associate Professor Zhang Mingxing, and others, cover key areas including high-performance computing, parallel and distributed systems, storage systems, intelligent computing systems, and LLM inference infrastructure. The integration of these outcomes marks a substantial step in the industry-academia-research collaboration between Approaching.AI and Tsinghua's research community in the field of AI infrastructure.
Within this framework, Academician Zheng Weimin, Chief Scientific Advisor at Approaching.AI, laid the academic foundation for high-performance computing at Tsinghua. Professor Wu Yongwei, Chief Scientist at Approaching.AI, has long focused on distributed and storage systems and has received numerous national-level science and technology awards. Associate Professor Zhang Mingxing specializes in LLM inference architecture and leads the development of widely adopted open-source projects such as KTransformers and Mooncake. Leveraging the injection of core Tsinghua technologies and the ongoing support of a top-tier research team, Approaching.AI has established a significant systems engineering and research-to-product barrier in AI inference infrastructure.
This technical strength is also validated in the open-source ecosystem. KTransformers, co-led open-source by Approaching.AI and the Tsinghua team, is the world's first edge-side heterogeneous inference framework and has garnered over 17,000 GitHub stars. In distributed inference, Approaching.AI collaborates with Tsinghua University and other industry-academia-research partners on the open-source Mooncake project. Yang Ke, a Tsinghua Ph.D. and technical expert at Approaching.AI, along with other team members, serves as a core contributor, deeply involved in the implementation and architecture of key technologies. Additionally, Approaching.AI actively contributes to the global inference community, continuously fostering the development of an open ecosystem for AI inference infrastructure.
The commercial team's grasp of industry demand, customer scenarios, and capital pathways, combined with the technical team's long-standing expertise in high-performance computing, distributed systems, and LLM inference infrastructure, equips Approaching.AI with the complete capability to deliver everything from low-level systems R&D to enterprise-scale implementation. As ATaaS continues to evolve, this multidisciplinary team structure will remain a cornerstone, supporting the company in enhancing the large-scale production and delivery of high-quality Tokens.
City:Beijing, China
Company Name: Approaching.AI
Website:https://www.approaching-ai.com/
Contact Person: Qu Xin
Email: info@approaching-ai.com
Report this content
If you believe this article contains misleading, harmful, or spam content, please let us know.