Halfway through his keynote at REAL Summit 2026 in Seoul on September 8, Samsung SDS CEO Lee Jun-hee named a number that does not sound like robotics news: 430.
That is how many manufacturing clients Samsung SDS has built manufacturing execution systems and equipment-and-logistics automation for over 25 years, concentrated in semiconductors, secondary batteries and biopharma. Then he named the next thing. Starting in 2027, the company will ship a robot orchestration platform that connects production equipment and robots and runs them as one system.
Either sentence alone is unremarkable. Together they are the story.
We covered the first half of this on July 25, when Samsung Electronics folded its scattered robotics teams into the RX division with a ₩19 trillion budget and a stated goal of turning every Samsung plant into an “AI autonomous factory” by 2030. September 8 supplies the second half. The Samsung company that sells software for a living has now said the capability is not staying in-house.
What the platform is actually for
Samsung SDS did not pitch a smarter robot. It pitched traffic control.
The platform’s stated core function is to compute collision-free movement paths for machines doing different jobs (packing, assembly, material transport), from different vendors, running on different cycle times. Its second function matters more. It watches for breakdowns in real time and reassigns the failed machine’s tasks to other robots, so one dead unit does not stall the line behind it.
Read those two functions against the last 18 months of humanoid coverage, which has been almost entirely single-unit metrics. Xiaomi named a 98.0% success rate on one nut-fastening station in its interim results. Unitree reports shipment counts. Hyundai walked JPMorgan through 25,000 Atlas units on May 19. Every one of those numbers answers the same question: can one machine finish one task?
Samsung SDS is answering a different one. Who schedules the fifty or two hundred mixed machines already on the floor?
That is the question that has been quietly gating deployment. A 98% station is a solved station. A plant is a scheduling problem, and until now the scheduler has been a person with a whiteboard and 15 years of shop-floor memory. Robot vendors have spent three years competing on the first thing. Nobody has sold the second.
One layer down, there is a prerequisite worth naming. Orchestration only runs if every machine reports state: what it is doing, which step it is on, how long is left, whether it is about to fail. Today that telemetry sits inside each vendor’s proprietary stack and does not cross. The first thing Samsung SDS has to do is collect it. Once it has, it becomes the only party that knows a plant’s real capacity.
The other two announcements in the room
The summit drew more than 15,000 participants on-site and online, and robotics was one track of several. Samsung SDS also said it is joining OpenAI’s Daybreak Partner Program as a pilot partner — a cybersecurity initiative that gives a small set of verified partners access to OpenAI’s most advanced models to probe for vulnerabilities. And in July it became the first Korean company to sign a strategic partnership with Anthropic.
The booth told the same story as the platform. Samsung SDS exhibited robots from Rainbow Robotics, its own affiliate, alongside hardware from Wuzi, a Chinese robot-hand specialist. It is not planning to sell only Samsung metal. It is selling the dispatch layer above the metal, and whose robot plugs in underneath is somebody else’s problem.
That is a deliberate place to stand. Robot hardware is getting cheaper fast, the vendor list is getting longer, and the machines are drifting toward commodity. What stays scarce is the knowledge of how a specific plant actually runs, and that knowledge has been sitting inside the MES for 25 years.
Team REX, and who is in it
Samsung SDS launched the effort with a partner bloc called Team REX, short for Robotics Excellence, bringing together 10 global physical-AI companies. Two named publicly are Rainbow Robotics, the KAIST-origin firm Samsung controls, and Walden Robotics.
Walden is the Toyota Research Institute alumni company valued at $1.1B. We wrote about it on July 16: wheeled base, no legs, robots already working shifts inside a Toyota plant. Samsung SDS has now signed a strategic investment and cooperation agreement with it, and the proof-of-concept work is specific rather than aspirational: Walden’s general-purpose robots handling parts and swapping consumable components on live production lines.
Walden CEO Russ Tedrake put the division of labor plainly in a video message: physical AI will reshape manufacturing, but turning it into measurable results on a real floor takes large-scale field deployment capability and business understanding, not just technology.
Translated: robot companies cannot get themselves into factories. Somebody has to walk them in.
The company walking them in installed the MES in those factories 25 years ago.
The sequence is in-house first, then resale
Joseph Ahn, EVP of Samsung SDS’s Digital Factory Business Unit, laid out the order: start with automation demand at Samsung’s own manufacturing sites, verify general-purpose robot performance there, then consider expanding into global markets.
Korean press reporting puts the first wave at roughly 1,000 worksites across major affiliates including Samsung Electronics, Samsung Heavy Industries and Samsung E&A. The 430 external MES customers come after that.
The sequence lines up exactly with July. We wrote then that Samsung’s own line workers are not the market for the product; they are its data set. September 8 closes the loop. Prove it on your own floors, then push it down a distribution pipe you have been laying since 2001.
Lee also opened with a McKinsey finding: 88% of global businesses have deployed at least one AI system, and 6% say AI is making a meaningful contribution to profit. He used the gap to frame the whole keynote. Adoption is not the same as outcome.
The gap is sharper on the robotics side than on the software side. Buying robots does not produce throughput. The missing piece is the scheduler.
Which jobs sit on this line
The first roles exposed are not line operators.
They are the production planners, shift supervisors, equipment dispatchers and maintenance schedulers. In a plant today, the calls of which machine runs which order first, where work goes when Line A stops, and who covers for a downed unit are made by people, working from floor knowledge and a hand-maintained board. Samsung SDS’s feature list is that job, itemized: path planning, collision avoidance, real-time fault detection, automatic reassignment.
The second change is distribution. Most of those 430 clients will never build a robot and will never write a scheduler. What they buy is software that plugs into the MES they already run. That turns substitution capacity from something only the largest manufacturers can self-build into a line item on a purchase order. And the three industries Samsung SDS named share clean lines, standardized process steps, and labor costs that are already high: semiconductors, secondary batteries, biopharma. That is the shortlist of environments where general-purpose robots land first.
The timing needs stating clearly. The platform ships in 2027, runs inside Samsung first, and reaches external customers later than that. Nobody loses a job next year because of this announcement. What was published on September 8 is a schedule.
Korean unions read schedules. On July 20, 35,000 Hyundai workers escalated to four-hour strikes, and the sticking point was not pay — it was a demand that no robot enter a plant without a prior labor agreement. The machine they were bargaining against was Atlas, which arrives in 2028. The date Samsung SDS just published is closer than that.
Sources
- Samsung SDS Enters Robot Transformation Business With 2027 Platform — Seoul Economic Daily, 2026-09-08
- Samsung SDS outlines AI transformation biz strategy at REAL Summit — The Korea Times, 2026-09-08
- Samsung Backs Walden’s Mission to Put Robots to Work Today — Walden Robotics, 2026-07-28
