Introduction
A September analysis of public procurement records says Unitree products appeared in 55 procurement orders and 89 tender lots in August, roughly one third of 168 disclosed embodied-robot projects. For medium and full-size humanoids and quadrupeds, Unitree was associated with 136 units.
Those numbers show strong institutional demand, but the destination of the spending matters. Research, education and data-collection centers dominate the disclosed projects, while direct industrial and field-use projects remain a smaller share.
Source review date recorded for this article September 24, 2026.
Research demand dominates the tender mix
The compiled data puts total August embodied-robot tenders around RMB 550 million. Research and education flows account for a large share, while data-collection and training centers account for much of the spending value.
ChainCatcher, citing GG Robot statistics, reports 168 disclosed projects, 55 procurement orders containing Unitree products and 89 related lots. It also reports that data-collection and training centers represented about 60% of disclosed spending and research or education nearly 80% of projects.
That means tender volume should not be read as an equivalent number of factory deployments. A university lab buying robots for research has a different utilization profile from an industrial site purchasing robots for repetitive work.
Unit counts are easier to compare when body classes are separated
The 136-unit figure focuses on medium/full-size humanoids and quadrupeds. Including desktop humanoids changes company rankings because low-cost education platforms can appear in larger quantities.
A market tracker should therefore keep body size, robot type and end use in separate columns.
Commercial demand needs follow-through data
The next useful data is acceptance and operating use after procurement. Track units delivered, installed, actively used and hours accumulated in the stated application.
A Tongji University notice provides a concrete research example: one Unitree G1 EDU U2 with a RMB 180,000 budget, a 10-day delivery requirement after contract activation and payment after acceptance. One university order should not be generalized to the full August market, but it shows how the aggregated tender count includes research purchases.
That follow-through would show how much procurement becomes routine work instead of equipment sitting in training centers or labs.
A procurement award is not the same stage as a robot in routine service
A public tender may describe a buyer's intended robot model and budget, yet several steps remain before hardware is delivered and used. Contract execution, equipment acceptance, technician training and integration with a research or production workflow can all occur later. Counting orders for university or data-collection laboratories as functioning industrial deployments would distort the interpretation of the market.
When comparing the reported August dataset, keep the bid announcement, award, signed contract, delivered units and accepted equipment as distinct events whenever the original records permit it. Also distinguish small educational models from full-size humanoids and quadrupeds. A single university procurement illustrates the structure of the records; it cannot establish how many hours the broader purchased fleet accumulated after delivery.
Sources and methodology
Share this article
Share the current TechniaHQService article page.
Continue reading
Open the latest robotics reporting, Physical AI analysis and hardware notes.