---
title: "Qianxun's RMB 4 billion funding sprint bets the robot 'brain' matters more than the body"
date: 2026-11-15
category: Capital & Markets
site: NeuroAI
canonical: https://neuroai.site/a/na-capital-qianxun-intelligence-funding
language: en
---

# Qianxun's RMB 4 billion funding sprint bets the robot 'brain' matters more than the body

> Qianxun raised over RMB 4 billion in 2026 rounds, crossing a RMB 10 billion valuation on its embodied-AI models.

Inside a CATL battery plant, a robot has placed nearly a thousand cells without a single fault. The startup that built its "brain" has just collected more cash in six months than most robotics firms raise in a lifetime — a bet that software, not steel, decides who wins the humanoid race.

Qianxun Intelligence (千寻智能), the Hangzhou embodied-AI (具身智能) company founded in 2024 by robotics veteran Han Fengtao (韩峰涛) and AI researcher Gao Yang (高阳), has become one of 2026's most heavily funded names. It closed two rounds totaling nearly RMB 2 billion (≈US$278M) in February, added about RMB 1 billion (≈US$139M) in a March–April round led by Shunwei Capital (顺为资本) and Yunfeng Fund (云锋基金), and followed with a roughly RMB 1.5 billion (≈US$208M) A+ round in May. That puts its first-half haul above RMB 4 billion (≈US$556M) and cumulative funding past RMB 5 billion (≈US$694M), with a valuation above RMB 10 billion (≈US$1.39B).

## Who is writing the cheques

The investor roster reads like a map of China's industrial economy. Beyond the financial names — Yunfeng (linked to Alibaba's Jack Ma), Shunwei (linked to Xiaomi's Lei Jun), Sequoia China (红杉中国), Chaos Investment (混沌投资, Ge Weidong) — the shareholder list includes manufacturers CATL (宁德时代), Inovance (汇川技术) and TCL, consumer-electronics giants Huawei (华为) and Xiaomi (小米), and logistics-finance players JD.com (京东) and China Merchants Venture (招商局创投). Local government funds from Chongqing and Hangzhou also took part.

That mix is the point. Qianxun is not selling a robot body; it is selling the general-purpose "embodied brain" (具身大脑) — the models that let a machine perceive, plan and act across tasks — and it has lined up the customers who would deploy it.

The pace itself is the signal. Qianxun took in roughly RMB 2.5 billion (≈US$348M) across the two March–May rounds alone. The rounds suggest capital is betting that embodied intelligence has moved from demo to deployment. The risk such speed creates is a valuation set by the last round, not by delivered margin — a gap that widens the longer commercial proof lags the fundraising.

## The technical claim

The company open-sourced its Spirit v1.5 model in January 2026 and has since advanced to v1.6. Both emphasize zero-shot generalisation — the ability to perform tasks like wiping objects or handling hinges and flexible items without new training samples. Architecturally, Qianxun blends a vision–language–action (VLA) model with a world model for action and future-state prediction, arguing that training on vast internet video is more efficient than predicting every frame.

Data is the moat it cites. Qianxun says it has gathered more than 200,000 hours of real interaction data through wearable capture devices — at about a tenth of traditional collection cost — and targets 1,000,000 hours in 2026. The volume, the founders argue, matters less than diversity: messy, real scenes teach generalisation better than clean, curated ones.

## Where it meets the factory

The commercial proof is a production line. Qianxun says its "Xiaomo" (小墨) robot runs on CATL's Zhongzhou base, placing nearly 1,000 battery cells with zero faults at a pace matching or beating skilled workers. A May partnership with Bosch (博世) extends the data-and-deployment loop into Bosch's Chinese factories, logistics centers and supply chain. The company frames 2026 as the year embodied-model performance breaks out, anchored on data scale.

## The risk in the round

Money this concentrated invites a correction. Qianxun's own backers note the sector is splitting between body makers — Unitree (宇树科技), AgiBot (智元机器人), UBTECH — racing to ship hardware, and brain-focused firms like Qianxun and Galaxy General (银河通用) betting on models and software–hardware decoupling. The danger for the latter is that a great model still needs a body that scales, and bodies are where delivery risk lives.

The bet is not unique. Galaxy General and other model-first shops are pursuing the same decoupling, so Qianxun's edge rests on data diversity and deployment speed rather than a proprietary architecture. A RMB 10 billion (≈US$1.39B) paper value therefore prices a race the company has not yet won, only entered with unusually deep pockets.

According to Li An, chief scientist at BrainNet (脑机网), a RMB 10 billion (≈US$1.39B) valuation on a model company is really a wager that data volume converts to generalisation before a hardware rival locks in the installed base.

## Honest limitations

Round sizes, dates, lead investors and the above-RMB 10 billion (≈US$1.39B) valuation are drawn from China Securities Journal, China Securities News, Hangzhou Daily and a Haidian District government release; no single prospectus consolidates them, and some figures (e.g., a later A+ near RMB 1.5 billion) appear in trade press rather than filings. The 200,000-hour data figure and 1,000,000-hour 2026 target are the company's own statements. The CATL line and Bosch partnership are reported by the company and local media; "zero-fault" placement is a company claim, not an independently audited metric. Comparisons to other embodied-AI firms describe strategy, not verified financials. Nothing here is a recommendation to invest in any round.

## What readers can do now

- Read the cap table as a customer list. When CATL, Huawei, Xiaomi and JD appear as backers, the investment and the offtake intention are often the same decision.

- Watch the data target, not the headline raise. The 1,000,000-hour 2026 goal is the variable most likely to separate a real generalisation lead from a funding story.

- Separate brain from body. A model company's value depends on whether its software rides many bodies or just its own — track deployment breadth, not just total funding.

---

Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-capital-qianxun-intelligence-funding
Free to quote with attribution and a link to the original.
