The Digital Twin Lineage — NASA, GE, Siemens

"A digital twin consists of a physical asset, a data model, and the information flow that connects the two." — Michael Grieves, U. Michigan (2002)

The digital twin is often introduced as "the latest technology," but the concept's roots go back more than 20 years. NASA operated backup craft on the ground as "physical twin" models from the Apollo era, and in 2002 Professor Michael Grieves of the University of Michigan formalized the term "Digital Twin" in the PLM field.

In the twenty-odd years since, GE Predix, Siemens Xcelerator, and NVIDIA Omniverse have each commercialized the digital twin from a different perspective, redrawing the industrial map. Understand this lineage, and the design direction of your "next digital twin project" changes.

The digital twin is not a technology — it is "a way of rewriting the questions we ask of physical assets."

1. NASA and Michael Grieves — The Birth of the Concept

NASA is recorded as the first organization to operate a "twin model of a physical asset," used to recreate conditions from the ground during mission control for Apollo 13 (1970). At a PLM industry conference in 2002, Professor Grieves gave the concept its name, arguing that all the knowledge that can be generated should be consolidated into a single virtual model.

In the 2010s, NASA formalized this in its "Modeling, Simulation, Information Technology & Processing" roadmap and applied it to the operation of the Space Shuttle, Mars probes, and the ISS. The idea of opening a data window onto physically inaccessible assets was later transplanted into industry — and that is the story of everything that followed.

Apollo era — the "physical twin" model operated on the ground (the prototype)
Grieves 2002 — the term "Digital Twin" formalized in the PLM field
NASA 2010s — officially incorporated into the MSITP roadmap
Three-element definition — physical asset + data model + bidirectional information flow
Industrial transplantation — commercialized through the combination of PLM, CAE, and IIoT

2. GE Predix — Commercializing the Industrial Asset DT

GE declared the "Industrial Internet" in 2012 and placed the digital twin at the heart of its commercial services. GE Digital's Predix platform provides Asset Performance Management (APM) services for power plants, refineries, and airline engines. The cumulative number of product twins exceeds 1.2 million.

Predix's other name — and its technical core — is SmartSignal. Using "peer group" models trained on lifetime data from hundreds of equipment types, it identifies early warning signs like fingerprints. With the 2024 spin-off of the energy business as GE Vernova, Predix became more clearly positioned as a specialized digital twin platform for the energy and utilities sector.

GE Predix was the first case of commercializing "digital twins of individual asset lifecycles."

3. Siemens Xcelerator + NVIDIA Omniverse — The Digital Twin in the AI Era

Since 2023, Siemens and NVIDIA have been integrating Siemens Xcelerator and NVIDIA Omniverse under an Industrial Metaverse strategic partnership. This alliance has produced the following technology market indicators.

Siemens share of the high-end DT deployment market: about 24% (industry #1)
GE Digital Predix cumulative product twins: 1.2 million+
Siemens 2025 automotive assembly line process twin cycle-time reduction: 15–25%
Siemens EV manufacturing demo: energy efficiency improvement of up to 22%
Siemens Digital Twin Composer announcement: CES 2026 (early access)

4. Case in Point — the Siemens Demo at NVIDIA GTC 2025

At NVIDIA GTC 2025, Siemens unveiled a demo that captures the entire lifecycle "from design to operations" in a single digital twin on the Xcelerator + Omniverse stack. AI automatically generates and simulates hundreds of factory layouts, and once the plant is physically deployed, the same model connects to real-time sensor data. The 3D model from the design phase becomes the master dashboard of the operations phase.

Siemens also performs "synthetic data generation" for autonomous robot training on top of the digital twin, generating unlimited rare samples in the digital environment for safety-part inspection and deformation prediction. This is the point where the "physical twin" of the NASA era evolved into a "data-generating asset."

How PlantPulse Answers

In a market where Siemens, GE, and NVIDIA offer industry-specific commercial DT platforms, KOPENS PlantPulse takes charge not of "DT features" but of "the industrial data operations layer that underpins the DT." It connects the ISA-95 asset model, real-time time-series data, lineage tracking, and the event hub to DT engines such as Siemens Xcelerator, GE Predix, and NVIDIA Omniverse through standard APIs.

The advantage of this structure is clear. Customers are not locked into a single DT vendor, and they can choose the DT platform best suited to each process (e.g., Predix for power plants, Xcelerator for new line design, Omniverse for autonomous robot training) while keeping the underlying data layer unified.

Closing

Looking at the digital twin's lineage makes one thing clear: this is not a story of "a new technology arriving." From the physical twin of the Apollo era, through Grieves's theory, Predix's commercialization, and the AI layer of Xcelerator + Omniverse — each stage has simply expanded the answer to the question "what can we replace with data?"

Over the next five years, the digital twin moves beyond the "visible factory" to the "deciding factory." It is the era in which physical assets and AI begin to "converse" on top of the digital twin. (References: Grieves "Digital Twin: Manufacturing Excellence Through Virtual Factory Replication" 2014, NASA MSITP Roadmap, GE Digital Predix APM, Siemens-NVIDIA Press "Industrial Tech Stack for AI-Era Manufacturing" 2025, Siemens Xcelerator Community "Industry Signals 2025/06/10", CES 2026 Siemens Digital Twin Composer)