Moscow Startup Summit Showcases a Robot That Is Learning to Work on the Factory Floor

In late September 2026, SberCity hosted the Moscow Startup Summit, organized by the Moscow Government and Sber. At the startup expo, visitors were greeted by a robot that makes eye contact, moves its arms and talks. Its tasks are down to earth: it is learning to perform specific operations on a production line. Habr author Sergey Lavrov described it in his second report from the summit. The published announcement does not name the model, the developer or any measurable specifications, so what follows covers only confirmed facts and a framework for evaluating them.

What we know about the expo

Lavrov's first report covered the talks at the summit. The second works like a walk through the exhibition floor: the author shows how science-fiction ideas, from I, Robot to The Martian, are turning into working technology. The headline promises several such examples, but the announcement reveals only one, a humanoid robot.

According to the description, the machine interacts with people almost like a movie character: it holds eye contact, gestures and answers aloud. The author is quick to add that the robot has no plans to save humanity. Its current job is to learn to do a handful of very concrete things on a production line.

• Event: Moscow Startup Summit, late September 2026.
• Venue: SberCity, Moscow.
• Organizers: the Moscow Government and Sber.
• Format: a program of talks and a separate startup expo.
• Exhibit from the announcement: a robot with voice interaction, movable arms and eye contact.
• Purpose: production tasks; training is ongoing.
• Not disclosed: developer, model, payload capacity, speed, autonomy, price, pilot stage.

Why the focus on production

At trade shows, humanoid robots almost always win on spectacle. Making eye contact and holding a conversation play well with a crowd, but that is the easiest part of a demo. Value on a factory floor is determined by other things: how many operations per shift the robot completes without operator intervention, how often it makes mistakes and how much it costs to maintain.

The phrase "learning to perform tasks" honestly signals the stage. This is training, not serial operation. For the robotics market, that stage is typical: the company has a prototype, a demo scenario and one or two pilot inquiries from industry. An expo where startups show up alongside city and corporate customers is a good place to look for such pilots.

The summit's format is also telling. When a city government and the country's largest bank co-organize an event, startups get a short path to first pilots and to contacts with corporate buyers. How well it actually worked for the exhibitors will become clear from pilot announcements over the coming months.

How to evaluate demos like this

Until more details emerge, it helps to keep a short list of questions in mind that applies to any robot at an expo:

• Who is the developer, and how many prototypes already work outside the booth?
• Which operations exactly does it perform: carrying, sorting, assembly, quality control?
• What is the success rate over a series of hundreds of repetitions, rather than a single show?
• Does it work autonomously or with a teleoperator, and where does the human sit in the loop?
• What does ownership cost: price, maintenance, integration with the line?
• Is there a named customer and a pilot timeline?

If you run a production site and are eyeing solutions like this, go into a pilot with metrics set in advance: cycle time, defect rate, number of stoppages per shift. The answers to these questions will tell you more about the machine than any video from an expo. A detailed breakdown of the exhibits will make sense after reading the full report on Habr.