Qujing Technology completed its Series A financing, with the total financing amount exceeding 1 billion yuan within half a year.
Recently, Approaching.AI, a high-efficiency AI Token production service provider, officially announced the completion of its Series A financing. Within half a year, the company's total accumulated financing amount has exceeded 1 billion yuan.
This round of financing is led by Henan Investment Group Huirong Fund, with oversubscribed follow-on investments from existing shareholders including Xinglian Capital, Zhenzhi Capital, Shangshi Capital, Shanghai Guofang Innovation, Honghui Fund, Huakong Fund, and Hangzhou Fucheng. The raised funds will be mainly used to expand the production capacity reserve of high-quality AI Tokens, upgrade the self-developed high-efficiency AI Token production service platform ATaaS (Approaching.AI Token as a Service), promote the large-scale deployment of domestic heterogeneous computing power in core production scenarios, and further build high-quality AI Token factories targeting leading large models, internet platforms, and regional industrial ecosystems.
The significant capital injection from the lead investor of this round values far more than ordinary financial returns, but the industrial prospects that the two parties can create together. Currently, the two sides are jointly advancing the preparation and construction of a high-quality AI Token factory with a daily production capacity of trillions of tokens.
For Approaching.AI, this financing marks that its high-quality AI Token production capacity has rapidly transitioned from single-point verification to the stage of large-scale supply. Since the 2026 Spring Festival, the company's average AI Token production efficiency per unit of computing power has increased by more than 3 times, and the total high-quality AI Token production capacity has grown by over 30 times; among them, for a leading large model with trillion-level parameters, the daily output of high-quality AI Tokens has steadily exceeded the trillion-token mark.
This rapid growth does not rely on simply stacking computing power or widely increasing the number of models, but stems from the company's continuous advancement of system engineering optimization relying on leading technological innovation capabilities in high-load and high-complexity real production environments. On this basis, Approaching.AI has formed a complete and replicable full-process closed-loop model covering the design, construction, production, and operation of high-quality AI Token factories. At present, the company has both completed and put into operation projects with a daily production capacity of trillions of tokens, as well as strategic cooperation and construction plans targeting leading customers and regional industrial ecosystems, promoting AI Token factories to gradually develop from single-point projects into scalable and sustainably operable new AI infrastructure.
From "Total Computing Power" to "Efficiency and Quality"
As large model applications evolve from pilot exploration to large-scale industrial implementation, the core evaluation criteria for AI infrastructure in the market have been iterated and upgraded. In the early stage of industry development, the industry focused more on basic resource capabilities, such as computing power supply scale, number of connected models, and unified access interfaces. However, when AI is deeply integrated into the core production links of enterprises, industry demands have gone beyond the construction idea of simply piling up computing power resources and stacking model quantities. What enterprises truly need is high-quality AI Token service capabilities that can support business implementation in a long-term, stable, and efficient manner.
This means that the core value of AI infrastructure is shifting from "total computing power supply" to "high-quality AI Token production capacity supply".
Approaching.AI defines "high-quality AI Token" as the smallest production unit that connects computing power, models, and application results. An AI Token that truly has enterprise-level implementation value needs to support models with hundreds of billions or even trillions of parameters, while taking into account low first-token response latency, stable high output speed, high concurrency processing capability, continuous output quality, reliable structured output and function calling, and controllable unit cost.
It is not the most difficult to achieve any single one of these capabilities. The real challenge is that all these indicators need to be met simultaneously under real production loads and remain stable during long-term operation. According to data calculations, different combinations of capabilities may result in production efficiency gaps of several times or even dozens of times. Only when computing power can be continuously, efficiently, and predictably transformed into high-quality AI Token production capacity that balances performance, quality, stability, and cost can it truly constitute a sustainably operable AI infrastructure.
Based on this judgment, Approaching.AI proactively proposed the concept of Token as a Service (TaaS), and through its self-developed ATaaS platform, deeply integrates underlying computing power and model inference systems downward, and seamlessly connects to the real business scenarios of enterprises upward, committed to building high-quality AI Token factories oriented to real production loads.
Compared with MaaS services that focus on computing power supply and model access, the ATaaS platform solves problems such as model performance, inference efficiency, resource utilization, cache reuse, service isolation, elastic scaling, quality monitoring, and cost control through full-link system engineering capabilities, thereby realizing large-scale production and stable delivery of high-quality AI Tokens.
"Fewer Models, Deeper Optimization" Focuses on AI Token Production Efficiency
The success of a high-quality AI Token factory ultimately depends on two core propositions: whether the per-unit computing power efficiency can be continuously improved, and whether the large-scale supply capacity can grow stably.
Focusing on this goal, Approaching.AI has established the technical route of "Fewer Models, Deeper Optimization", concentrating on AI Token demands in enterprise production scenarios to help customers maximize the return on investment of AI.
Approaching.AI does not take the number of models and throughput scale as the sole goal, but focuses on two variables: First, continuously and deeply optimize a small number of large models with real production demands, and increase the number of effective AI Tokens that can be produced per unit of computing power through model segmentation, video memory management, and heterogeneous collaboration. Second, improve the stability and predictability of production loads and reduce production capacity loss and computing power idleness through enterprise-exclusive AI Token services, cross-cluster resource coordination, fault recovery, and AI Token factory operation.
In other words, "Fewer Models, Deeper Optimization" solves the problem of "how many AI Tokens can be produced with the same batch of computing power", while stable production load and cross-cluster operation solve the problem of "how much time these computing power units are actually in effective working state". "Fewer Models" does not naturally bring higher utilization rates. Only when both per-unit computing power output and effective utilization rate are improved can the unit economics of the AI Token business be viable.
The business logic behind this route is very clear: Enterprise-level customers ultimately pay for business results, not for the number of compatible models.
At present, Approaching.AI has formed a replicable high-quality AI Token factory construction and operation model: there are completed projects with a daily production capacity of trillions of tokens, as well as strategic signing and construction plans for regional industrial ecosystems and leading customers. More importantly, some mature businesses have already crossed the cost threshold.
Enabling domestic computing power to support the production of high-quality AI Tokens has always been Approaching.AI's goal. To this end, the company has invested a large amount of R&D resources, and successively proposed a series of technical solutions such as "Domestic Prefill-Decode (PD) Heterogeneous Collaboration", "High-Performance Heterogeneous KVCache Conversion", and "Heterogeneous Computing Power Pooling", which have been officially put into production in high-standard production scenarios, providing practical and effective implementation solutions for domestic chips to achieve high-quality AI Token production.
Deep Closed-Loop of Industry, University, Research, and Application, Building the Trust Foundation of the Industry
Business Closed-Loop: Compound Management Team Drives Industrial Implementation
Crossing the gap from scientific research, open source, engineering implementation to commercial landing, the company has built a talent team with strong transformation potential: the management team led by Chairman Ren Xuyang, CEO Dr. Ai Zhiyuan, President Dr. Wu Wenjie, and CTO Chen Xianglin has fully opened up the management closed-loop of technology R&D, capital operation, and industrial application. At present, Approaching.AI has transformed its leading underlying technology and systematic engineering capabilities into the ATaaS platform that can be delivered in a standardized manner, and has taken the lead in running through the deployment path for leading customers.
Technical Origin: Profound Scientific Research Heritage Supports Long-Term Original Innovation
Behind the steady commercialization process lies the profound scientific research heritage and technical soil from the High-Performance Computing Institute of the Department of Computer Science, Tsinghua University. The technical team is supported by top experts: Academician Zheng Weimin of the Chinese Academy of Engineering serves as the chief scientific consultant, Professor Wu Yongwei of Tsinghua University serves as the chief scientist, and Associate Professor Zhang Mingxing from the Department of Computer Science of Tsinghua University and co-founder of Approaching.AI has long led the company's technical strategy and key R&D breakthroughs, continuously promoting cutting-edge technological innovations. Based on this, Approaching.AI, based on the original innovation of underlying system software technology, has breakthroughtly proposed industry-first concepts such as "full-system heterogeneous collaboration", "calculation through storage", and "virtual-real isomorphism", contributing an important "Chinese solution" to solving the computing power challenges of the artificial intelligence industry.
Open Source Ecosystem: In-Depth Co-Construction Builds Industry Technical Trust
While consolidating underlying technologies, Approaching.AI deeply participates in the evolution of key technologies for large model inference infrastructure. KTransformers, an open source project led by the company and the Tsinghua team, has received widespread attention in the industry. At the same time, the company has joined hands with Tsinghua University, Moonshot AI, 9#AISoft, Alibaba Cloud, Ant Group and other institutions to co-build the open source project Mooncake, and remains active in mainstream global AI inference communities such as vLLM and SGLang. These solid ecological practices have accumulated a technical trust foundation for Approaching.AI's subsequent commercial delivery.
From the open source ecosystem to enterprise-level production delivery, Approaching.AI relies on its built "high-quality AI Token factory" to directly transform technical barriers into substantial optimization of computing power costs and comprehensive improvement of system stability, laying a solid foundation for the large-scale growth of customers' businesses. Facing the accelerated implementation of agents, AI Coding, multi-modal applications, and enterprise-level AI workflows, Approaching.AI is becoming a key support for AI to move into deeper development with its stable, efficient, and predictable AI Token supply capacity.
Industry-Investment Resonance, Tapping the Long-Term Value of High-Quality AI Token Production and Service Capabilities
In the past six months, the capital market has continuously increased its investment in Approaching.AI, accelerating the financing process, with total raised funds exceeding 1 billion yuan. Relying on its industry-leading core technology, mature AI Token commercialization achievements, and broad industrial prospects, the company has gained high consensus and continuous recognition from the capital market. Multiple rounds of financing have attracted a large number of leading investment institutions to actively participate, resulting in over-subscription.
This round of financing was completed at a critical stage when Approaching.AI's commercialization is accelerating. As large model applications enter real production scenarios, the demand for high-quality AI Token production capacity from leading models, internet platforms, and regional industrial customers is growing rapidly. The new funds will further support the company to expand computing power reserves, upgrade the capabilities of the ATaaS platform, and promote the large-scale application of domestic heterogeneous computing power in core inference scenarios.
Dr. Ai Zhiyuan, CEO of Approaching.AI, stated:
"As large models are fully integrated into production systems, high-quality AI Token services with continuous stability, fast response, and controllable costs have become the core rigid demand for enterprises to implement AI on a large scale. Approaching.AI adheres to the technical route of 'Fewer Models, Deeper Optimization', continuously tackles the world's most cutting-edge technologies, pursues the ultimate conversion efficiency from computing power to high-quality AI Tokens, and builds world-class AI underlying service capabilities. This round of financing will further accelerate the large-scale implementation of Approaching.AI's high-quality AI Token factories. Through deep collaboration with industrial investors, we will promote the large-scale commercialization of domestic PD heterogeneous solutions, help domestic chips achieve large-scale and normalized production in high-standard AI production scenarios, and contribute an important Chinese solution to the evolution of global artificial intelligence."
Notably, the lead investor of this round has a profound industrial background. The two sides have based on strategic synergy and jointly launched the preparation and construction of a modern high-quality AI Token factory. The significant and in-depth participation of industrial investors reflects the capital market's recognition of the long-term value of AI Token factories, and also reflects the growing attention from the industrial side to high-quality AI Token production and service capabilities.
A person in charge of Henan Investment Group Huirong Fund, the lead investor of this round, stated:
"As large models accelerate their penetration into the application end, the demand for large-scale inference continues to be released. Efficient and stable high-quality AI Token supply has become the core infrastructure supporting the in-depth development of the AI industry, and efficient computing power conversion capability will be the core focus of the next stage of AI competition. Approaching.AI originates from the top high-performance computing team of Tsinghua University. With the hard power of full-stack software and hardware optimization, it has opened up the value chain from underlying resources to model output. Its ATaaS platform has achieved a trillion-level daily call volume, and is a key node connecting the computing power infrastructure and upper-layer applications. Huirong Fund is highly optimistic about Approaching.AI's key ecological position in the computing power industry chain and the core technical barriers it has built in the field of inference optimization. Relying on Henan Investment Group's profound industrial accumulation in the fields of green energy and computing power infrastructure, we will fully link the upstream and downstream industrial ecological resources, cooperate deeply with Approaching.AI, and help it grow into the world's leading high-efficiency AI Token production service provider."