Siemens Brings 1,000x-Faster AI Simulation to Casting — PhysicsAI Now Inside Simcenter Inspire

AI & Analytics · Simulation & Digital Tools · Europe/Global · August 2026

On 28 July 2026, Siemens announced availability of the latest Simcenter software release — the first to combine Siemens' and Altair's engineering simulation portfolios into a unified environment. The headline capability for manufacturing engineers is the expanded integration of Simcenter PhysicsAI across the portfolio, including directly into Simcenter Inspire — the simulation tool used by casting, moulding, forming, and extrusion process engineers.

The release is three weeks old and represents a material shift in how AI-accelerated simulation is delivered to manufacturing process engineers — not as a standalone research tool, but embedded inside the workflow environment they already use.

"Engineering simulation isn't limited by physics — it's limited by how quickly we can explore possibilities."

>

— **Sam Mahalingam**, Executive Vice President, Simulation, HPC and AI, Siemens Digital Industries Software

What Simcenter PhysicsAI Is

Simcenter PhysicsAI is Siemens' geometric deep learning technology — a surrogate modelling engine that trains on existing simulation data (historical solver runs, Design of Experiments studies, prior CFD results) and then predicts outcomes on new geometries without running a full solver.

The architecture uses a transformer neural network optimised for geometric data, allowing it to handle complex 3D shapes, varying topologies, and rich field outputs — not just scalar performance numbers. Once trained, the model predicts results directly on mesh or CAD geometry in seconds, with the high-fidelity solver retained as the validation reference.

The stated performance benchmark: up to 1,000x faster than traditional solver-based simulation. Real-world deployment data from Kinetic Vision (packaging design) showed the tool running at up to 4,000x faster than traditional FEA at 97.5% accuracy — suggesting the 1,000x figure is a conservative floor, not a ceiling, once surrogate models are well-trained on production data.

What's New in This Release — Four Capabilities

1. PhysicsAI integrated into Simcenter Inspire

The most relevant development for casting engineers. Simcenter Inspire already contains Inspire Cast — a dedicated metal casting simulation module covering gravity, high-pressure, low-pressure, and other casting processes. The PhysicsAI integration now enables AI-powered prediction within that casting workflow, enabling seamless data exchange between casting process simulation and AI-powered surrogate prediction. Engineers can run a parametric sweep of gate locations, wall thicknesses, or shot profiles in seconds rather than hours.

2. Simcenter PhysicsAI Generate — new generative capability

A new module introduced alongside the casting integration. PhysicsAI Generate is a physics-aware generative AI engine that creates novel design concepts based on target dimensions, performance KPIs, and historical training data — in seconds. This shifts the tool from pure prediction (evaluate a given design) to generative design (propose a new one). For casting component design, this opens the possibility of AI-proposed geometry that is optimised for fill behaviour and porosity risk from the outset.

3. Multi-GPU support in Simcenter STAR-CCM+

AI model training from CFD simulation data now runs significantly faster with multi-GPU support — achieving 10x speed improvements and beyond on training cycles. This reduces the time investment required to build a capable surrogate model from a company's existing simulation library.

4. Extended PhysicsAI in Simcenter Hypermesh

More efficient data training workflows and improved result confidence tools for structural simulation — relevant for die and tool designers working on die casting machine tooling.

The Altair Integration — Why It Matters

This release is the first commercial output of Siemens' acquisition of Altair's engineering simulation portfolio. The combination brings Altair's structural and multiphysics simulation capabilities (including tools well-established in the casting and stamping supplier base) into the Simcenter environment under a unified licence and workflow structure.

For casting houses and OEM simulation teams that currently operate mixed Siemens/Altair toolsets, this unification reduces tool complexity and enables connected workflows across disciplines — structural, thermal, fluid, and manufacturing process — within a single environment.

Competitors in this space include Ansys, Cadence, Dassault Systèmes (SIMULIA), Hexagon, and Autodesk — all of whom have introduced AI-assisted simulation and surrogate modelling technologies. Siemens' latest release distinguishes itself by integrating the Siemens and Altair portfolios into a unified environment while extending AI specifically across manufacturing process workflows.

What This Means for Pressure Die Casting Engineers

The direct application to HPDC and LPDC process engineering:

Gate and runner design iteration

Traditional casting simulation runs — fill analysis, thermal gradient prediction, air entrapment risk — can take hours per variant. A trained PhysicsAI surrogate model reduces each iteration to seconds, enabling engineers to evaluate dozens of gate configurations in a single session before committing to hard tooling.

Porosity risk screening

The combination of Inspire Cast's physics-based fill simulation and PhysicsAI's surrogate prediction creates a two-tier workflow: fast AI screening across a wide design space, followed by targeted high-fidelity solver runs on the shortlisted candidates. This is the same workflow logic as the Tohoku/Astemo research tool covered elsewhere in this issue — but delivered as a commercial, production-ready platform.

Earlier design freeze

By moving accurate process simulation to the concept stage — before die design is finalised — casting engineers can influence component geometry for manufacturability, not just validate it after the fact. This is the structural shift PhysicsAI enables: simulation as a design input rather than a design check.

Training on your own production data

The surrogate model trains on a company's own simulation history — prior DOE studies, historical solver runs, production part results. Casting houses with years of Simcenter or STAR-CCM+ data already hold the training asset. The barrier to adoption is lower than it appears.

Positioning Alongside the Tohoku/Astemo Research (This Issue)

This issue carries two simulation stories that address the same fundamental problem — predicting porosity and fill behaviour in HPDC before metal is poured — from opposite ends of the technology spectrum:

The two approaches are complementary, not competing. High-fidelity physics-based tools like DiecastCompressibleInterFoam generate the ground-truth simulation data that AI surrogate tools like PhysicsAI train on. The academic research and the commercial platform are parts of the same maturing simulation ecosystem.

Key Contacts

*Siemens Digital Industries Software is headquartered in Plano, Texas, with European operations in Munich. Simcenter product information at siemens.com/simcenter. PhysicsAI product page: siemens.com/simcenter/physicsai*

*Sources: PR Newswire (28 July 2026) · Siemens news.siemens.com · HPCwire · Engineering.com · Robotics & Automation News · Siemens Simcenter blog (whats-new-in-simcenter-physicsai-2026-1) · Siemens PhysicsAI product page · Industry EMEA*

*© 2026 Velocla · velocla.com*