---
title: "AgiBot's 10,000th robot rolled off the line in March, and its commercial ledger now reads like a parts list for other people's factories"
date: 2026-11-18
category: Companies & Stack
site: NeuroAI
canonical: https://neuroai.site/a/na-corp-agibot-commercial-scale
language: en
---

# AgiBot's 10,000th robot rolled off the line in March, and its commercial ledger now reads like a parts list for other people's factories

> AgiBot crossed 10,000 cumulative units in March 2026, with near-1,000-unit orders from Longcheer, a 100-million-yuan (about US$14 million) auto-parts contract from Joyson, and a dexterous-hand subsidiary valued above $1 billion — but valuation reports conflict and IPO timing is unclear.

The useful question about Chinese humanoid robots is no longer "can they walk" but "who is paying, and for what". AgiBot (智元机器人) is one of the few companies whose answer can be traced through named customers and contract sizes rather than demo videos.

In late March 2026, the company's 10,000th general-purpose embodied robot came off its production line. What happened on other companies' assembly lines around the same time tells you where the revenue actually sits.

## Key takeaways

- AgiBot's cumulative production passed 10,000 units in March 2026, which trade reports describe as the largest delivery scale among Chinese embodied-robot makers (CNR-sourced reports carried by Tencent News, 31 March 2026).

- Contract manufacturer Longcheer (龙旗科技) has placed orders for close to 1,000 units for 3C electronics assembly; auto-parts supplier Joyson Electronics (均胜电子) signed a contract worth over 100 million yuan (about US$14 million) for automotive-parts work.

- On Joyson's Ningbo line, AgiBot's G2 robot handles tri-pin flexible assembly at a cycle time as fast as 12.97 seconds with a success rate above 99% (company-disclosed figures, March 2026).

- Its dexterous-hand subsidiary, Lingjiedian (临界点, "critical point"), has passed a US$1 billion valuation, per multiple industry reviews.

- Valuation and listing details are murky: reports cite a parent valuation above 15 billion yuan (about US$2.1 billion) with a planned Hong Kong IPO in 2026, while market chatter about a new strategic round at a much higher figure remains unconfirmed — and reported figures differ between sources.

## From 10,000 units to named customers

The 10,000-unit milestone was announced on 31 March 2026. Company executive Wang Chuang framed it as a supply-chain rather than production-line achievement: going from a thousand units to ten thousand, he said, depended on the overall maturing of suppliers and quality systems. Peng Zhihui, the company's co-founder, tied the milestone to an "internet of machines" argument — that fleets of robots working in real scenes generate training data that makes the next robot more capable on arrival.

The more concrete evidence is in customers' own factories. At Longcheer, a contract manufacturer that assembles tablets for major brands, AgiBot's G2 works the MMIT (multimedia integration testing) station, doing precision loading and unloading. At Joyson's auto-parts plant in Ningbo, the same model performs tri-pin positioning flexible assembly — a task the companies describe as high-difficulty — with the disclosed cycle time and success rate above.

Trade reports reviewing the commercialisation of Chinese embodied-robot companies place AgiBot in the "delivery chain established" tier: contracts signed, units shipped, but repeat-purchase depth and overseas revenue share still unproven. Its customer base skews toward 3C contract manufacturing and auto parts, and its product line spans wheeled and bipedal platforms — more than seven models, per trade reports.

## The software layer and the subsidiary

Two pieces of the story sit outside the assembly line.

First, the models. AgiBot's GO-1 foundation model uses a ViLLA architecture with a mixture-of-experts (MoE) design, pairing latent action representations with expert routing. A separate model, WITA-Omni, is described in industry reports as the first embodied-interaction model in China to complete compliant filing (合规备案) — a regulatory status claim that comes from reporting rather than an independent registry check.

Second, the components. Its dexterous-hand subsidiary Lingjiedian has passed a US$1 billion valuation according to several industry reviews published this November. Dexterous hands are one of the scarce components in humanoid supply chains, and a separately capitalised subsidiary lets the company sell the part to competitors while shipping whole robots to customers.

## The money question, stated honestly

The financial picture is where sourcing gets uncomfortable. Reports circulated in mid-November describe AgiBot as having begun a new strategic financing round at a valuation around 50 billion yuan (about US$6.9 billion), with its listing pace deliberately slowed. Other reports — including broker and trade reviews — cite a parent valuation above 15 billion yuan (about US$2.1 billion) and a plan for a Hong Kong IPO in 2026. These two sets of figures cannot both describe the same thing in the same terms, and the higher figure has not been confirmed by any filing this review could locate. What can be said with confidence: the company is reported to be raising again, its listing timeline is described as flexible, and the valuation range in circulation is wide.

That matters because the commercial ledger, while real, is young. Near-1,000-unit orders from one contract manufacturer and a nine-figure-yuan contract from one auto-parts supplier are meaningful, but both come from manufacturing-adjacent customers where robots do repetitive physical tasks. Whether the next tier of customers renews — and whether revenue moves from project-based deployments to recurring contracts — is the question trade analyses keep flagging.

## Honest limitations

Most figures in this article come from company disclosures relayed through Chinese industry and financial media, not audited filings. The 10,000-unit count is cumulative production, not cumulative deployed units or paid deliveries. The Joyson cycle time and success rate are company-stated performance figures from a single station on a single line. The Longcheer order size ("close to 1,000 units") is approximate by construction. The GO-1 architecture description and the WITA-Omni compliance-filing claim both come from secondary reporting. And the valuation figures conflict across sources, as stated above — readers should treat any specific number attached to a new financing round as unconfirmed until a transaction closes.

## What readers can do now

- If you evaluate robot vendors, use the disclosed metrics as interview questions: ask any supplier for cycle time and first-pass success rate on your own part geometry, not a showcase station.

- If you track the sector, watch for the two signals trade reviews identify as next-stage markers: repeat orders from existing customers, and overseas revenue as a share of deliveries.

- If you follow the capital side, treat valuation chatter as noise until a filing appears; the verifiable datapoints are the HK IPO plan reported for 2026 and whatever terms the next round actually closes at.

---

Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-corp-agibot-commercial-scale
Free to quote with attribution and a link to the original.
