Ubtech Robotics Balances Equity Incentives, Fenglong Integration and a Cheaper Route to Robot Training

Ubtech Robotics has had a busy week on several fronts at once: it has issued fresh H-shares to the trustee of its incentive scheme, closed the acquisition of Fenglong Stock, and unveiled a simulation platform designed to cut the cost of teaching intelligent machines. The stock, however, slipped 3.0% on the day the Thinker-Sim product news landed.

Incentive shares and a completed takeover

The company confirmed it issued 1,958,700 new H-shares to the trustee of its H-share incentive programme. The shares cover awards already granted to directors and employees. Of the total, 59,700 shares are earmarked for three directors, while 1,899,000 shares are destined for 360 staff members.

On Thursday, Ubtech said — according to media reports — that it had wrapped up its acquisition of Fenglong Stock. The subsidiary founded by Fenglong, Xin Fenglong, has moved two-finger grippers and servo drives for bionic faces into production. According to the company, it intends to develop and manufacture hardware for embodied artificial intelligence.

Thinker-Sim: a loop between factory floor and simulation

The newly presented Thinker-Sim platform rests on a Real2Sim2Real approach. Its Thinker-GS component reconstructs real scenes, such as factory lines, from which simulation data can then be generated and robot functions tested. Those checks span perception, navigation and object handling.

Ubtech argues the approach trims the cost of gathering data, the effort of building simulations and the time needed to validate algorithms. Training in virtual environments, the company adds, should also lower the expense and risk of physical trials. The payoff therefore lies not merely in more data but in a more systematic preparation for real-world tasks.

Management puts the rendering efficiency gain at 67.3%, and says templates and no-code visual workflows cover more than 90% of target scenarios. Those figures describe technical performance characteristics of the platform, not a corresponding rise in revenue or profit. Investors must therefore separate the promised efficiency gain in robot training from its eventual economic effect.

Analysts see data pipelines as the next battleground

Guosheng Securities analysts expect data collection to grow in importance for physical artificial intelligence models. They anticipate combined data pipelines drawing on teleoperation, motion capture, simulation and data gathering without a robotic body, alongside an expansion of open datasets.

Thinker-Sim picks up part of that trend with synthetic data and standardised tests, tying the generation of training material directly to the assessment of capabilities. That takes the approach beyond a mere digital replica of an environment.

Growth provides the commercial backdrop

Ubtech doubled its revenue in the first half of 2026 compared with the prior-year period, reaching 1.27 billion yuan, while the number of humanoid robots sold climbed to 3.7 times the previous level. On 28 August, the company had already reported that revenue from products and solutions featuring human-sized humanoid robots with embodied intelligence jumped 1,445.0% year on year to RMB590.3 million.

That growth gives the software initiative a concrete corporate context: Ubtech is expanding its development tools in step with rising robot sales. Simulation and evaluation are meant to make the path from training data to practically deployable capabilities more efficient.

Demonstrations, with a caveat

On Friday, Ubtech disclosed six tests pairing GPT-6 Astra with its robots. Under direct control, one attempt to grasp a cube failed to complete the task. Another test combined Astra with Ubtech's Thinker-VLA and mastered a separate task involving the grasping of various objects. The company stressed, however, that the demonstrations did not constitute a rigorous comparison on the same task.

Thinker-Sim offers a technical route to lower development effort. Whether that translates into commercial advantage will be settled in practical use — not by the efficiency metrics alone.