CNPC and Unitree plan a joint venture for industrial robots

On November 4, the Beijing Municipal Administration for Market Regulation published a simplified merger filing: CNPC Kunlun Manufacturing, together with its concert parties Kunlun Capital, Unitree Robotics and Beijing Fanjie Intelligent Technology, plan to establish a joint venture engaged in industrial embodied-AI robot solutions. The public comment period runs to November 13.

The split is 60, 20, 20

After the transaction, Kunlun Manufacturing and its concert parties will hold 60%, with Unitree and Fanjie at 20% each; the three parties will jointly control the venture. The structure says two things: the CNPC side wants control, and Unitree takes a minority stake while supplying technology.

The business scope is specific: "industrial embodied-AI robot solutions." Not selling robot bodies — selling solutions. In state-owned procurement the difference is large: a solution bundles selection, integration, commissioning and maintenance, often including changes to existing lines.

Petrochemicals is not virgin territory for robots. Pipeline crawlers, explosion-proof inspection vehicles and automated loading systems have run for years. What humanoids and quadrupeds are chasing is the work that needs stairs, doors and valves — jobs that stayed with people because the machine's shape was wrong.

What each side brings

CNPC brings sites and capital. Inspection, hauling, valve operation and work in hazardous areas are chronic labour shortages that still have to be done; Kunlun Manufacturing builds equipment, Kunlun Capital invests — an industry-plus-capital combination.

Unitree brings the body and locomotion control. Its experience with quadrupeds and humanoids maps onto petrochemical needs: stairs, pipe racks, narrow spaces.

Fanjie is the least documented of the three; public reports do not describe its background. Judging by the business scope, it most likely handles integration or delivery — every venture of this kind needs someone responsible for putting the solution on site.

Why a "solution" rather than "a batch of robots"

State-owned enterprises buy equipment through a fixed process: approval, tender, acceptance, audit. A single robot struggles to pass that process because it cannot produce independently verifiable returns; a solution can be packaged as a project with clear deliverables and acceptance criteria.

That also explains why a separate company is needed instead of Unitree simply supplying equipment. A joint venture can take projects from inside the group, keeping R&D, integration and maintenance under one roof, with risk and liability contained in a single entity.

For Unitree it is a stable order channel; for CNPC it is a controllable testing ground. Both sides get something, which makes the relationship firmer than a purchase order.

Joint ventures between state-owned groups and tech companies are not new. Over the past decade, their fate has usually turned on the first project: whether it passes internal acceptance and copies to a second plant. One demonstration leads to replication; a failure leaves a company with a name and no business.

A filing is not a deal

To be clear, this is a simplified merger notification under antitrust rules, not a contract announcement. The comment period runs to November 13 and any interested party may object.

Several things remain undisclosed: registered capital, which sites or oilfields come first, whether the arrangement is exclusive, and what Unitree contributes (cash, technology licences or equipment). Those details decide whether this is real investment or a placeholder.

For Unitree there is another layer: having listed this year, it needs steady revenue to support its growth story, and a state-owned order book is the best endorsement available.

The trend is nonetheless clear: state-owned enterprises are using joint ventures to bring robotics into their own industrial systems rather than waiting for suppliers to mature. For robot companies, orders may increasingly come not from open tenders but from the internal procurement of a company they partly own — a steadier path, and a more relationship-dependent one.

This piece summarizes public reporting and did not independently verify vendor demos or experimental data. Not investment advice.