On 6 January 2026 at CES in Las Vegas, Siemens introduced Digital Twin Composer. Strip away the event packaging and what remains is an old practice, repackaged: do not change a live production line before you know the consequences.
What the tool connects
Digital Twin Composer takes a 3D digital twin built with the Siemens Xcelerator portfolio and connects it to real plant data: manufacturing execution software (MES), quality management systems (QMS), the PLC code of individual machines, and IIoT data. The rendering side uses NVIDIA Omniverse libraries.
The connection is the point. A beautiful 3D model with no link to operational data is an animation. Once it reads real machine state and real control logic, it becomes a place to experiment: change a conveyor layout, reorder a process step, add a machine, and see what breaks before anyone picks up a wrench.
Siemens says the tool is in early access with selected customers and is expected on the Xcelerator Marketplace in mid-2026. In the same announcement the company set out plans for nine AI copilots across software including Teamcenter, Polarion and Opcenter.
The numbers quoted
The case Siemens cites is PepsiCo, recreating every machine, conveyor, pallet route and operator path with physics-level accuracy. The reported results: up to 90% of potential issues identified before any physical modification, a 20% throughput improvement in the initial rollout, and a 10–15% reduction in capital expenditure.
These are vendor figures presented at the vendor's own event, about an organisation with resources unlike most plants in Vietnam. Read them as evidence that the approach works, not as the level of return to plan around.
For enterprises in Vietnam
Most plants in Vietnam neither need nor should buy a digital twin platform yet. But the discipline underneath applies immediately, and costs almost nothing: before changing a running process, be able to answer "what if this is wrong" with data rather than instinct.
At its simplest that means running in parallel. The new process runs alongside the old one for a few weeks, results are compared, and only then does the switch happen. A step up is simulating on historical data: take six months of output and incident records, run the new rule against that set, and see what decisions it would have made.
The common obstacle is not the missing tool but the missing data to simulate with. The plant has PLCs recording everything that nobody can extract, an MES used only to print tickets, and an incident spreadsheet living on one person's laptop. Before you can simulate anything, those sources have to reach one place a machine can read.
That is why most projects should start from the data layer and one small application with real users, not from a 3D model. A digital twin is a reasonable destination, but it only earns its keep once real data is flowing underneath it.
Sources
Siemens Press — Siemens unveils technologies to accelerate the industrial AI revolution at CES 2026 (6 January 2026): https://press.siemens.com/global/en/pressrelease/siemens-unveils-technologies-accelerate-industrial-ai-revolution-ces-2026
Siemens News — Siemens unveils Digital Twin Composer: https://news.siemens.com/en-us/digital-twin-composer-ces-2026/