NVIDIA introduced the Apollo family of open models for accelerating industrial and computational engineering, unveiled at the SC25 conference in St. Louis. These physics-optimized models—developed for scalability, performance, and accuracy—will enable developers to integrate real-time capabilities into simulation software across fields including structural mechanics, weather and climate, and computational fluid dynamics. Industry leaders such as Applied Materials, Cadence, and Siemens intend to train, fine-tune, and deploy their AI technologies utilizing these new open models, with Applied Materials achieving up to 35x acceleration in modules of its ACE+ multi-physics software through NVIDIA GPUs and the CUDA framework.
NVIDIA Apollo: Open Models for Scientific Simulation
NVIDIA Apollo is a new family of open models designed to accelerate industrial and computational engineering, unveiled at SC25. These physics-optimized models are intended for diverse fields including electronic device automation, structural mechanics, weather/climate forecasting, computational fluid dynamics, electromagnetics, and multiphysics. The models utilize advanced machine learning architectures like neural operators, transformers, and diffusion methods, combined with domain-specific knowledge, and will be available as pretrained checkpoints and reference workflows.
Several industry leaders, including Applied Materials, Cadence, and Siemens, are adopting NVIDIA Apollo to enhance their technologies. Applied Materials is leveraging NVIDIA AI physics to improve semiconductor manufacturing power efficiency, achieving up to 35x acceleration in modules of its ACE+ software. Cadence created a real-time digital twin of a full aircraft using data from simulations accelerated by the NVIDIA-powered Millennium M2000 Supercomputer.
Synopsys is also utilizing NVIDIA AI physics to achieve significant speedups – up to 500x – in computational engineering. Initializing fluid simulations with AI physics surrogates, rather than traditional methods, significantly reduces runtime for GPU-accelerated tools like Ansys Fluent. NVIDIA Apollo models will soon be available via build.nvidia.com, HuggingFace, and as NVIDIA NIM microservices.
Applications Across Industries and Engineering Fields
NVIDIA Apollo open models are being adopted across multiple industries, including automotive, aerospace, and energy, to accelerate design processes. Companies like Applied Materials, Cadence, and Siemens are integrating these AI physics models into their existing workflows. Applied Materials, for example, has achieved up to 35x acceleration in modules of its ACE+ software, improving power efficiency in semiconductor manufacturing. This highlights the potential for significant performance gains by blending traditional simulation with AI surrogates.
Several companies are leveraging NVIDIA Apollo for specific engineering challenges. Cadence utilized NVIDIA’s Millennium M2000 Supercomputer to create a dataset for training an AI physics model enabling a real-time digital twin of a full aircraft. Similarly, Northrop Grumman is using NVIDIA CUDA-X libraries and Luminary Cloud’s platform to rapidly explore spacecraft thruster nozzle designs. KLA intends to use the models to accelerate simulations for semiconductor process control solutions.
Significant speedups are being realized through the integration of NVIDIA Apollo models. Synopsys is achieving up to 500x speedups in computational engineering, while Siemens is exploring designs orders of magnitude faster than previously possible. Rescale is also enhancing its AI physics operating system with these models to allow engineers to blend high-fidelity simulations with high-speed AI surrogates, maintaining accuracy while dramatically reducing development cycles.
AI Physics Implementation and Acceleration Results
NVIDIA Apollo is a family of open models designed to accelerate industrial and computational engineering through AI physics. Several industry leaders—including Applied Materials, Cadence, and Siemens—are adopting these models to improve their design processes across automotive, aerospace, and other fields. Applied Materials has already seen up to 35x acceleration in modules of its ACE+ multi-physics software by utilizing NVIDIA GPUs and the CUDA framework, enabling faster semiconductor process optimization.
Synopsys is leveraging NVIDIA AI physics to achieve significant speedups—up to 500x—in computational engineering applications. This acceleration is achieved by combining GPU acceleration with AI physics surrogates, drastically reducing simulation runtimes for tools like Ansys Fluent. Cadence also developed a real-time digital twin of a full aircraft by training an AI physics model using a dataset generated with its Fidelity Charles Solver, accelerated by the NVIDIA-powered Millennium M2000 Supercomputer.
Rescale is integrating NVIDIA Apollo models into its AI physics operating system to allow engineers to blend traditional, high-fidelity simulations with high-speed AI surrogates. This combination enables the exploration of vast design spaces much faster, achieving real-time inference results while maintaining the accuracy of conventional simulation methods. Siemens is taking a similar approach by integrating NVIDIA AI physics into its Simcenter STAR-CCM+ fluid simulation tools.
