Investing 13% of revenue in R&D across RISC-V architecture, Wi-Fi 7, Matter protocol, and on-device AI — turning lab breakthroughs into silicon you can ship.
Each pillar is staffed by a dedicated team with end-to-end ownership from research to tape-out.
Designing purpose-built RISC-V processors optimized for IoT workloads: ultra-low power wake, single-cycle GPIO access, and integrated DSP extensions for audio and sensor fusion.
Active development of 802.11be MLO (Multi-Link Operation) silicon enabling simultaneous 2.4/5/6 GHz band utilization with 4096-QAM modulation for sub-millisecond IoT latency.
Core contributor to the Connectivity Standards Alliance. Our Thread 1.3 and Matter 1.3 implementations are among the first certified stacks shipping in production silicon.
Neural network accelerator IP capable of 100+ GOPS at under 50 mW — enabling keyword detection, anomaly classification, and gesture recognition without cloud round-trips.
Secure boot chain, flash encryption engine, hardware TRNG, and dedicated cryptographic coprocessors (AES-256, RSA-4096, ECDSA) integrated at the silicon level.
Research into sub-threshold voltage radio architectures targeting < 5 µA deep-sleep with instant-on wake — extending battery-powered IoT sensor life beyond 7 years.
Where we have been and where we are heading.
First production RISC-V-based wireless SoC with integrated Wi-Fi 6 and BLE 5.0, replacing legacy ARM cores in our IoT lineup.
Among the first chipset vendors to achieve full Matter certification — enabling interoperability across Apple HomeKit, Google Home, and Amazon Alexa ecosystems.
Integrated neural network coprocessor shipping in production, enabling real-time keyword spotting and gesture recognition at the edge without cloud dependency.
Next-generation SoC with 802.11be Multi-Link Operation across 2.4/5/6 GHz bands — targeting sub-1 ms deterministic latency for industrial and automotive IoT.
We believe the best wireless technology is built collaboratively.
Our complete development framework is open-source on GitHub with 12,000+ stars, 4,500+ forks, and daily community contributions spanning 40+ languages.
500,000+ registered developers with active forums, regional meetups, annual DevCon conferences, and a partner certification program for system integrators.
Collaborative research programs with 8 leading universities focusing on low-power radio design, RISC-V extensions, and edge AI optimization.
No single wireless protocol fits every deployment. Understanding these trade-offs helps architects choose the right stack.
Wi-Fi proponents argue that leveraging existing IP-based infrastructure eliminates the need for dedicated gateways, simplifies cloud connectivity, and benefits from Wi-Fi 6 improvements in dense environments (OFDMA, TWT).
Thread/Zigbee proponents counter that mesh topologies provide superior reliability in large-scale sensor networks, consume significantly less power (enabling 5+ year battery life), and offer self-healing network resilience that Wi-Fi star topologies lack.
Our approach: we offer both protocol stacks on the same silicon, allowing customers to evaluate each in their specific RF environment before committing to a deployment architecture.
Proprietary RTOS supporters emphasize deterministic scheduling guarantees, long-term vendor support with SLAs, and pre-certified safety libraries critical for automotive (ISO 26262) and medical (IEC 62304) applications.
Open-source advocates point to faster innovation cycles, community-driven security auditing, elimination of per-unit licensing costs, and the flexibility to customize the kernel for specific power/performance profiles.
Our ESP-IDF is built on an open-source FreeRTOS foundation, which we supplement with commercial-grade security modules and certification support for regulated industries.
Transparency about what our platforms can and cannot do helps you make informed design decisions.
Our integrated antenna modules are optimized for short-to-medium range applications (typically 50-100 m line-of-sight for Wi-Fi, 10-30 m for BLE). Deployments requiring multi-kilometer range should consider our LoRa-compatible modules or external antenna designs — not our standard SoC modules.
On-device AI inference is limited to lightweight models (keyword spotting, simple image classification under 300 KB model size). Complex vision models, large language models, or real-time video analytics exceed the on-chip memory and compute budget — these workloads require edge gateway offloading or cloud processing.
Standard modules operate within -40 to +85 °C. Deployments in extreme environments (foundries, furnaces, cryogenic storage) exceeding this range require custom packaging and extended qualification testing, which adds 8-12 weeks to the development timeline and is available only through our custom module design service.
Whether you are a startup with a bold idea or an enterprise looking for a technology partner, our innovation lab is open for collaboration.
Request an Innovation Briefing Explore Our Platforms