Alsemy needed to replace slow, expert-only semiconductor device-modeling workflows that required months or years of manual physics fitting before chip teams could move on new process nodes.
Reported impact
Alsemy cut device-modeling cycles from years of physics work and months of manual fitting to weeks of GPU training, one-second inference, and 10-minute customer tuning, helping semiconductor R&D teams move faster with lower cost and expert-level accuracy.
Alsemy rebuilt semiconductor device modeling around physics-informed AI so customers no longer have to wait months or years for experts to hand-fit SPICE and T-CAD models. Its Alsis platform learns transistor behavior directly from I-V and C-V datasets, while Alsphere extends the approach to structural and process modeling, including a display-focused variant co-developed with LG Display. Model development runs on NVIDIA RTX 3090 GPUs with NVIDIA PhysicsNeMo and CUDA-accelerated PyTorch across a hybrid on-premises and cloud environment, letting a small team iterate quickly on pretrained Retarget and C-V models. Once trained, the models can be fine-tuned with customer data and then deployed for CPU-based inference inside customer environments, so fabs keep sensitive data on-premises while still benefiting from GPU-accelerated training and rapid experimentation.
I-V and C-V measurements plus real silicon data feed PhysicsNeMo and PyTorch training on RTX GPUs, pretrained and customer-specific models are refined in Alsis and Alsphere, and the resulting models are delivered for CPU-based inference inside customer R&D and fab environments.
Alsemy helps chipmakers replace year-long SPICE or T-CAD model development cycles with workflows measured in minutes or days.
Alsis models are typically deployed for CPU-based inference in customer environments.
The team trains large pretrained models, including a Retarget model and a C-V model, using NVIDIA RTX 3090 GPUs in a hybrid infrastructure.
a few weeks of GPU training, plus one-second inference and 10-minute customer-specific tuning
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Aug 25, 2026 Siemens Siemens needed to help industrial teams overcome fragmented engineering tools, labor shortages, and slow design and maintenance workflows so they could design, simulate, and operate complex products and factories faster… 10%–15% reductions in capital expenditure NVIDIA CUDA · NVIDIA DGX B200 · NVIDIA DGX H200 NVIDIA Sample Aug 24, 2026 Noble Machines Noble Machines needed to train, evaluate, and deploy a general-purpose industrial robot fast enough for real customer environments without relying on slow, physical-only testing or cloud-dependent inference. 15 people NVIDIA GeForce RTX 5090 GPUs · NVIDIA H100 GPUs · NVIDIA Isaac GR00T 1.7 NVIDIA Sample Aug 12, 2026 Pegatron Pegatron needed to improve factory planning, quality control, and operational efficiency across large manufacturing sites, but manual line design, fragmented agent development, scarce defect data, and labor-intensive… 10% decrease in operational effort NVIDIA AI Enterprise · NVIDIA Cosmos · NVIDIA DGX 