March 10, 2026 How GlobalFoundries and MIPS enables “sense–think–act–communicate” architectures for radar, SATCOM, and electromagnetic advantage. Ashish Shah, Deputy Director, Aerospace and Defense, GlobalFoundriesEric Schulte, Sales Director, MIPS Defense RF platforms are evolving from “sense-and-stream” architectures to “sense–think–act–communicate” Physical AI architectures where inference and control occur as close to the antenna as possible to shorten decision loops and improve performance in contested environments. This shift is driven by rapidly increasing spectrum and waveform complexity, paired with strict size, weight, power and cost (SWaP‑C) constraints, and the growing need for trusted microelectronics in sensitive defense applications. In practical terms, operators need RF systems that can adapt to dense spectral occupancy, interference, jamming/spoofing, multipath and complex multi-function sensor operations without relying on high latency backhaul to centralized compute. As spectrum becomes more crowded, and decisions more real time, defense systems must redesign RF architectures to achieve in-theatre electromagnetic advantage with closed-loop, edge-resident intelligence. Physical AI is not a software add-on with associated high overhead. It is a real-time workload within the RF signal chain driving performance upgrades to RF fidelity, compute, power/thermal design and mission assurance from silicon through deployment. What changes with Physical AI in RF systems (systems view) Traditional RF architectures often capture, digitize and stream data to downstream processing, resulting in limited ability to respond quickly to dynamic threats or changing propagation conditions. Physical AI changes this paradigm by placing decision-making in the loop, enabling the RF system to sense–think–act–communicate. For radar, SATCOM/communications and electromagnetic advantage, this Physical AI approach translates into energy efficient functions such as adaptive waveform selection, beam/mode scheduling, interference recognition/avoidance, emitter classification and spectral triage happening closer to the sensor—reducing latency and bandwidth demand while improving resilience. Three system-level challenges to address 1) RF fidelity (wide spectral bandwidth + dense interference) Physical AI implementations in RF applications are only as good as the signals they receive. If the RF front end saturates, distorts, or loses linearity in the presence of blockers, downstream feature extraction and inference can be compromised. System architects should continue to improve linearity, isolation and predictable RF behavior as they look to implement Physical AI. Example outcomes: Radar: higher signal-to-noise ratio improves AI-assisted clutter/interference handling and classification. SATCOM/Comms: stable front-end behavior improves interference avoidance and link adaptation decisions. Signal Intelligence: high isolation protects feature extraction for real-time classification and geolocation under co-site emitters. 2) SWaP‑C + thermal headroom for in-loop inference and control Implementing intelligence at this level adds compute and memory requirements near the RF sensor. Many defense platforms have tight power, thermal and size constraints while requiring deterministic timing. Embedded processors enable event-driven “compute bursts” (infer when needed, return to low-power monitoring when idle) and provide predictable control paths to keep the sensing loop stable while staying within constraints. 3) Uninterrupted supply of uncompromised microelectronics Readiness and deterrence require assured access to microelectronics components that are securely designed, manufactured and tested, with robust protections for confidentiality and integrity and verifiable provenance traceability. Thwarting the adversary also demands less vulnerable microelectronics that integrate more of the digital and RF functions. For adaptive systems, primes also need a credible path for secure updates (firmware and models) via secure communications. GF technologies support both CMOS and high-performance RF circuit integration to design advanced signal processing functions with high performance RF signal chains. GF and MIPS enable embedded Physical AI Physical AI success depends on more than inference throughput. It requires deterministic closed-loop control inside the RF chain. MIPS embedded cores best serve as that decision and control anchor that turns RF observations into timely actions, supporting local classification, policy selection and real-time modification of the RF signal chain. Integrated using GF technologies, this approach can reduce integration and adversary compromise risk (fewer off chip interfaces), improve timing predictability and support qualification paths that require trusted manufacturing options. MIPS anchors its value proposition with targeted solutions aligned to customer-specific workloads. This enables defense integrators to tune microarchitecture and real-time features to the exact sensing, classification and control loops required by mission profiles while simultaneously optimizing power and performance. By nature of the RISC-V instruction set architecture (ISA) being an open standard, customers can perform independent verification of the privilege models and security extensions while also avoiding vendor lock-in or any opaque microarchitectural behaviors. Additionally, the open standard of RISC-V enables tailored ISA extensions, hardware roots of trust and domain-specific cryptographic accelerators. Mission capability is advanced, and integration risks are mitigated by using MIPS Atlas Explorer virtual platform for digital engineering with software-first development. This digital twin of the CPU core provides early workload validation and pre-silicon performance modeling necessary for shortening development and qualification timelines. The entire Atlas portfolio from MIPS is purpose-built for Physical AI workloads by combining deterministic control, scalable compute and security primitives to support the next generation of RF systems. MIPS embedded Physical AI cores at the RF edge (Proven Partner): • Deterministic real-time control to close the sense→think→act→communicate loop near the RF data path • Configurable architecture to tailor compute to RF-control + lightweight inference workflows • Low-power integration aligned to SWaP‑C • Secure deployment alignment to support long-life platforms and controlled updates within trusted design, manufacturing, and test flows Building the right RF-to-AI architecture from the start If you are developing AI-assisted RF capabilities for radar, SATCOM/communications, or signal intelligence, GlobalFoundries can help map system requirements to the right combination of RF platform, help you architect embedded MIPS Physical AI cores and trusted secure supply options—accelerating from architecture definition to fielding tested qualified microelectronics while reducing system integration risk. Engage GF (MIPS) early to align on the embedded processor subsystem choices for deterministic Physical AI at the RF edge.
