Flexcompute is positioning its AIartificial intelligence platform as a way to validate chip and package designs against physics before tapeout. The company describes its approach as a Physics World Model that combines neural networks with first-principles solvers, so engineers and AIartificial intelligence agents can estimate how a design will behave before anything is built.
At the center of the platform is a set of physics solvers for electromagnetics, fluid dynamics, structures and heat. Flexcompute says its validated solvers are fast enough to fit into AIartificial intelligence-driven design loops, while still serving as a verification layer grounded in physical laws. More than 250 companies already use the technology across semiconductors, photonics, aerospace, automotive and defense.
Photonics is the strongest application area. Teams at AMD, Anduril, NVIDIA, Samsung Display, Coherent, TDK, Ayar Labs and Global Foundries use the company’s tools to connect foundry PDKprocess design kitA PDK provides the technology files and design rules needed to design chips for a specific process.s with EDA1. electronic design automationSoftware tools for designing, simulating, verifying and laying out electronic circuits.2. equipment data acquisitionSystem or method for collecting, storing, exchanging or analyzing data. flows. Optical I/O and co-packaged optics for AIartificial intelligence data centers are described as the fastest-growing part of that business.
Flexcompute is also focusing on advanced packaging, where heat, stress, warpage and reliability interact as chiplets, HBM1. human body modelESD test model that simulates a person discharging static electricity into a component.2. High Bandwidth MemoryHBM is a stacked DRAM architecture designed to provide very high memory bandwidth. and optics share one package. The company says its warpage Physics Twin can predict which of more than 800,000 solder points will fail under thermal load before production. The same modeling approach is being extended to thermal and signal-integrity problems.
PhotonForge 2.0 was released in October. The updated flow is meant to let engineers and agents work on one design from PDKprocess design kitA PDK provides the technology files and design rules needed to design chips for a specific process. and schematic through layout, multi-physics simulation, DRCdesign rule checkA verification step that checks whether layout geometries comply with fabrication rules., LVSlayout verification of schematicVerification that an integrated-circuit layout matches its intended schematic. and tapeout. Recent additions include physics-verified compact models and automatic Verilog-A export, so device physics continues into circuit simulation.
Flexcompute also highlighted a photonic design example built in an agentic workflow: an optical I/O chip with eight wavelength channels at 100 Gb/s each. The company said each demultiplexer filter showed about −18 dB crosstalk to its neighbor, yet a full-link simulation with all channels present revealed cumulative leakage that noticeably degraded the eye. The example is used to underline the need for full-system simulation rather than isolated block checks.
To reach production workflows, the company is extending physics verification into standard EDA1. electronic design automationSoftware tools for designing, simulating, verifying and laying out electronic circuits.2. equipment data acquisitionSystem or method for collecting, storing, exchanging or analyzing data. environments and offering enterprise features such as foundry PDKprocess design kitA PDK provides the technology files and design rules needed to design chips for a specific process. integration, EDA1. electronic design automationSoftware tools for designing, simulating, verifying and laying out electronic circuits.2. equipment data acquisitionSystem or method for collecting, storing, exchanging or analyzing data. connectors, version control and traceability to tapeout and measured data. For the most demanding packaging cases, Flexcompute works directly with customers to build a Physics Twin that engineers and AIartificial intelligence agents can interrogate before fabrication.