Infinigence AI has raised nearly 500 million yuan ($69 million) in a Series A+ funding round as the Chinese artificial intelligence company accelerates its transition from traditional agent infrastructure to what it calls "Agentic Infra," a native architecture designed for autonomous AI agents deployed across cloud and edge environments.
The financing was led by Zhuhai Technology Group and Fortune Capital's Yuanchuang Future Fund, with participation from Huiyuan Capital, Shangqi Capital and Honghui Fund.
Infinigence AI said it will use the funds to strengthen its software-hardware co-optimization capabilities, expand adoption of its cloud and terminal AI products, and accelerate development of its agent infrastructure platform. The company aims to support large-scale deployment of autonomous agents across both digital and physical environments.
Co-founder and Chief Executive Xia Lixue said a structural shift is under way in artificial intelligence as autonomous agents become central to real-world applications, replacing traditional model-driven workflows that rely on extensive "glue code" and manual engineering.
"The paradigm shift brought by Agentic AI isn't just a strategic opportunity—it is a calling of our era," Xia said. "With years of engineering experience in AI systems, Infinigence AI has completed its native intelligent agent infrastructure transformation with agility. Looking ahead, we will continue to optimize software and hardware in tandem and build the next generation of learnable and evolvable Agentic Infra."
The company's approach is built around an "AI Agent Infrastructure × (AI Cloud + Edge Intelligence)" architecture. In this model, cloud systems act as the compute backbone powering training, reasoning and orchestration, while edge devices serve as the embodiment and sensory interfaces through which agents interact with the physical world.
On the cloud side, Infinigence AI operates WuQiong AI Cloud, a platform providing standardized agent-services capabilities and full-stack artificial intelligence services. The company says the cloud manages more than 25,000 petaflops of computing nationwide, spanning 53 core data centers across 26 cities. The platform offers high-performance computing, native toolchain support and agent hosting services for enterprises across industries.
More than 100 companies—ranging from AI model developers to internet platforms and device manufacturers—are using its cloud infrastructure. Customers include Baichuan Intelligence, Kimi, Lenovo Group, Liepin, Li Auto, Lovart, Sand.ai, Shushukeji, Soul, VAST, New H3C, Zhipu AI, China Mobile and ZTE Terminal. Research partners include Shanghai AI Lab, Shanghai Algorithm Innovation Research Institute, Zhijiang Laboratory, Beijing Academy of Artificial Intelligence and Zhongguancun College.
On the terminal side, Infinigence AI has launched a full-stack edge intelligence solution encompassing edge models, edge software and proprietary inference IP. Its flagship multimodal edge model, Infini-Megrez (also known as WuQiong Tianquan), requires only 3 billion compute and 7 billion memory units while offering intelligence levels equivalent to 21 billion-parameter systems, addressing constraints typical of consumer and industrial hardware.
The company's terminal inference engine, Infini-Mizar, reduces latency threefold and cuts both energy consumption and memory usage by 40% on mainstream chips. Meanwhile, its self-developed inference LPU IP, Infini-Merak, aims to lower the cost of large-model inference while doubling energy efficiency.
In addition to infrastructure, Infinigence AI has introduced two agent-focused products:
• Agents Infra, a cloud-native agent swarm system supporting orchestration and autonomous collaboration; and
• Kernel Mind, a general-purpose inference acceleration platform for edge devices.
Supporting technologies include the RLinf reinforcement learning framework, which enables continuous evolution of agents, and the Cache-to-Cache communication protocol designed to improve efficiency and reduce latency in inter-agent communication.
Xia said the company is building an "AI-native infrastructure" where users of future AI applications will not need to understand which model or hardware accelerator is running in the background. "Before we turn on the tap, we don't need to know which river the water comes from," he said.
Zhuhai Technology Group's chief investment officer Liu Feihong said agentic systems are pushing AI toward a new stage defined by autonomous perception and decision-making, and that Infinigence AI has built a mature product matrix across cloud and edge environments.
"Leveraging the rich scenarios of intelligent manufacturing and smart cities in the Guangdong-Hong Kong-Macao Greater Bay Area, we will work with Infinigence AI to promote a closed loop of technology, industry and application in agentic systems," Liu said.
Yuanchuang Future Fund said the company has demonstrated strong technical foresight, citing its work in heterogeneous computing, cloud-edge integration and unified scheduling. Huiyuan Capital added that Infinigence AI's strategy aligns with China's long-term objectives to build a self-reliant AI ecosystem under the country's 15th Five-Year Plan.
Infinigence AI said the new capital will support broader commercialization of agent-based AI across manufacturing, automotive, consumer devices and smart-city infrastructure, positioning the company for the next phase of growth as agentic systems transition from experimentation to scaled deployment.
