Samsung Electronics has set up a platform to share semiconductor process data in real time with its materials, parts, and equipment suppliers — a step toward jointly analyzing fab data with partners, folding it into AI-based factory operations, and ultimately moving toward fully unmanned chip plants.

According to a June 16 report by ETNews, Samsung is running a platform called DSEP, for Data Sharing Eco Platform, with participation reportedly surpassing about 60 companies — led by equipment suppliers — and still growing. DSEP shares a portion of the process data Samsung had previously kept to itself, collecting and analyzing it on the platform and feeding it into AI models. The aim is to open up data whose use had been limited by security concerns, in order to sharpen defect detection and stabilize yield. The scope of shared data is bounded, but the move is a notable shift from a company that has long treated fab data as among its most closely held secrets.

Why a Secretive Company Is Opening Up

The change is practical, and it cuts against decades of habit. Key equipment data — error codes, processing times — has been hard to take off the factory premises for security reasons, so when a machine showed anomalies, the supplier's engineers had to come on-site to inspect it in person. Sharing and analyzing that data on the platform in real time lets both sides identify equipment status and the cause of problems quickly, without anyone being dispatched.

The reason Samsung is willing to loosen its grip is scale. As process nodes shrink, each step generates rapidly growing volumes of data — reportedly several billion data points at a single fab — and timely analysis becomes essential to catch equipment anomalies and quality variables early. That is more than any one company can comb through alone, and because suppliers understand their own components' structures and failure patterns intimately, analyzing the data jointly speeds up root-cause identification. The partners, in turn, get real production-line data to refine their own equipment and materials. It is a trade: some control over guarded data in exchange for faster, AI-assisted diagnosis.

The Endgame: A Factory That Runs Itself

Samsung plans to use AI models built on the platform to strengthen equipment inspection, failure prediction, and defect-probability analysis. The effort dovetails with its stated goal of converting its fabs to 100% unmanned operation by 2030 — an ambition for which the company has said building a collaborative ecosystem across the supply chain is essential. It has already laid groundwork, including a high-performance computing center within its DS division to support data analytics.

The logic of the lights-out fab is what ties it together. A factory can only run without people if it can sense equipment trouble and quality drift automatically and in real time, then act on it — which is impossible if the data needed to spot those problems is locked on-premises and read only after an engineer flies in. DSEP is the data-collection layer of that loop; the AI models are the analysis layer; and the goal is a closed cycle of collect, analyze, predict, and correct that needs no one on the floor.

Samsung has shown what automation can yield, though so far mainly in packaging rather than across a full fab. In its chip-packaging lines, the company says unmanned conversion has cut manufacturing headcount by about 85%, reduced equipment failures by roughly 90%, and more than doubled overall equipment efficiency — results from the portion of its packaging lines already converted, which illustrate the potential rather than the current state of its wafer fabs.

One Track Among Several

DSEP is not Samsung's only path toward the autonomous fab. The push runs parallel to its work with Nvidia on Omniverse-based digital twins of its fabs for anomaly detection and predictive maintenance — another track toward what both companies call a fully autonomous fab, and part of a broader sector race in which rivals are pursuing the same 2030 horizon.

An unmanned fab is not simply about cutting on-site staff; how far it can go depends on how reliably process data can be collected and analyzed. The steady expansion of DSEP's partners is read as a strategy to advance a data-driven operating system incrementally, which is why the platform's results are expected to become a key gauge of how fast Samsung can make the transition. As one industry official put it, AI adoption is spreading among equipment suppliers as they use DSEP's shared data, accelerating the AI transformation across the partner ecosystem.

Frequently Asked Questions

What is Samsung's DSEP?

DSEP, short for Data Sharing Eco Platform, is a platform Samsung Electronics built to share a portion of its semiconductor process data in real time with materials, parts, and equipment suppliers. Reportedly used by more than 60 companies and growing, it collects and analyzes that data and feeds it into AI models to improve defect detection, equipment monitoring, and yield.

What is a lights-out (unmanned) fab?

A lights-out or unmanned fab is a chip factory designed to run with little or no human presence on the production floor. Samsung has said it aims to convert its fabs to 100% unmanned operation by 2030. Reaching that requires automatically sensing equipment problems and quality variations and correcting them in real time — which is why collecting and analyzing process data is central to the goal.

Why does Samsung share fab data with suppliers?

Historically, equipment data such as error codes and processing times stayed on-premises for security, so a malfunction meant dispatching a supplier's engineers in person. Sharing that data on DSEP lets both sides diagnose problems remotely and quickly. Because suppliers know their own equipment's failure patterns best, joint analysis speeds root-cause identification, and the volume of data a modern fab generates is too large to analyze alone.

How much can fab automation improve operations?

Samsung says that on the portion of its packaging lines it has converted to unmanned operation, manufacturing headcount fell by about 85%, equipment failures dropped roughly 90%, and overall equipment efficiency more than doubled. Those figures come from packaging rather than full wafer fabs, so they illustrate the potential of automation rather than current fab-wide results.