Beijing-based Approaching.AI, also known as Qujing Technology, has raised more than RMB 1.3 billion — approximately $195 million — in a Series B financing round announced on October 26, 2026. The round was backed by a group of state-linked, financial and industrial investors including Zhejiang Finance Capital, Xiaoshan Capital, ICBC Investment, Zhongguancun Science City, Xinglian Capital, Qianjiang Century City, SenseTime and Wangsu Science & Technology. The company has now raised approximately RMB 2.4 billion during 2026, according to earlier reporting.
What Inference Is, and Why It Costs So Much
To understand this round, it helps to understand the two phases of an AI model's life. Training is the expensive, one-time process of teaching a model. Inference is what happens every time the model answers a question — and it happens millions of times a day, forever. As companies deploy AI assistants and autonomous agents that may make multiple model calls to complete a single task, inference has become one of the largest recurring operating expenses in enterprise technology.
Approaching.AI builds software and computing infrastructure designed to run large language models more efficiently. Its work focuses on computing utilization, memory management and token-generation efficiency — in practical terms, getting more answers out of the same hardware, at lower energy cost. The company also develops technology intended to improve performance across different processor architectures, including Chinese AI hardware.
Why This Round Is So Large
A Series B of nearly $200 million is enormous by any standard, and the investor list explains part of it. The round blends state-linked capital (Zhejiang Finance Capital, Zhongguancun Science City), a major bank's investment arm (ICBC Investment) and industrial strategics (SenseTime, Wangsu). That mix reflects how AI infrastructure is treated in China: not merely as a venture bet, but as industrial policy.
The strategic backdrop matters. Access to advanced foreign chips remains constrained for Chinese companies, which raises the value of any technology that squeezes more performance out of available hardware. If a company can cut inference costs by a meaningful percentage, every AI application built on top of it becomes cheaper to run — and in a market where compute is scarce, efficiency is a competitive weapon, not just a cost saving.
The Inference Economy Goes Global
Approaching.AI's round is part of a worldwide pattern. In the West, inference prices have fallen sharply through 2026 as model providers compete on efficiency, and startups across the stack — from chip designers to software optimizers — are raising capital to own a layer of the inference economy. The economic logic is the same everywhere: whoever controls the cost of running AI models controls the margins of the entire application layer above them.
What differs in China is the structure of the capital. Where Western inference startups typically raise from venture funds, Approaching.AI's round reads like a coalition of state, financial and industrial backers assembled around a national capability. That gives the company patient capital and domestic customers, but it also means its long-term advantage will depend on independently comparable cost, performance and reliability under commercial workloads — benchmarks that global customers, not just domestic ones, will eventually demand.
The Risks Behind the Headline
For all its size, the round leaves important questions unanswered. No valuation was disclosed, and the company's commercial traction — paying enterprise customers, deployment scale, measured cost savings — has not been publicly quantified in the same detail as the financing. Inference optimization is also a fiercely competitive field: model providers themselves keep making their models cheaper to run, which can compress the value of third-party optimization layers over time.
Investors in this round are betting that the optimization layer remains valuable even as models get more efficient — that the sheer volume of AI usage will grow faster than efficiency gains, keeping inference costs a first-order problem for years.
What It Means for Founders and Investors
Approaching.AI's round carries three lessons. First, the AI investment frontier has moved from "who builds the best model" to "who makes models affordable to run" — infrastructure economics now command the largest cheques. Second, in constrained hardware markets, efficiency technology is strategic technology, and capital follows accordingly. Third, founders building anywhere in the AI stack should watch inference costs as closely as model quality: the company that cuts a customer's AI bill by 30% has a sales pitch that writes itself.
For investors, the deal is a reminder that the biggest AI rounds of 2026 are no longer concentrated in the United States. The inference economy is global, and the capital forming around it is too.
Angeled.com connects founders with investors.
Tech Startups, "Startup Funding News Today, October 26, 2026: Ailytics, Approaching.AI, CASBOT, Volare & More" (October 26, 2026)
Source: uxc.news
UXC News, "Beijing AI infrastructure startup Approaching.AI" (October 26, 2026)
Read Next
General Translation Raises $27 Million Series A Led by a16z to Build Agent-First Localization Infrastructure
LangChain Raises $300 Million Led by CapitalG at a $3 Billion Valuation to Build the Agent Stack
Mount Thor Raises $45 Million to Run Cloud Fleets of Macs for AI Agents
OpenAI Reportedly Raises $15 Billion at a $1.25 Trillion Valuation
Q-CTRL Raises $290 Million Series C Led by Salesforce Ventures, Reaching Unicorn Status
SE3 Raises $30 Million Series A Led by Accel



