According to Embedded Computing Design, the 2025 embedded world North America Best-in-Show winners reveal a diverse landscape of innovation across 18 categories. Okika Devices won with their OTC2902A SoC Field Programmable Analog Array featuring 192 amplifier transistors and 10,000 FPGA gates, while ADL’s AI2500 delivered 157 TOPS of AI performance in a rugged, fanless design. Other notable winners included Nordic Semiconductor’s nRF54L15 ultra-low power wireless SoC with 1.5MB memory, Exascend’s PD5 U.2 SSD offering 61.44TB capacity and 14 GB/s reads, and STMicroelectronics’ groundbreaking LSM6DSV320X MEMS IMU with dual accelerometers and AI-powered motion detection. The awards highlight significant advances in edge AI, mixed-signal processing, and rugged computing across multiple industries.
Sponsored content — provided for informational and promotional purposes.
The Dawn of Physical AI
What we’re witnessing across these winners is the emergence of what I call “Physical AI” – systems where sensing, computation, and action converge in real-time physical environments. The MIPS I8500 processor explicitly targets this era, but the trend appears across multiple winners. STMicroelectronics’ dual-range accelerometer can detect everything from subtle movements to high-impact shocks while running AI models locally, representing a fundamental shift from simply collecting data to making intelligent decisions at the physical interface. This isn’t just about putting AI on devices – it’s about creating systems that interact intelligently with the physical world in ways previously impossible.
The Analog Computing Renaissance
Okika’s FPAA victory signals something profound: we’re reaching the limits of digital-only computing for certain applications. While everyone focuses on digital AI accelerators, the programmable analog array represents a different approach entirely. Analog computing excels at tasks like signal conditioning, filtering, and certain types of mathematical operations that are inherently inefficient in digital domains. As we push AI into more real-world applications dealing with sensor data, analog preprocessing and computation may become essential for power efficiency and latency reduction. This could spark a broader revival of mixed-signal architectures optimized for specific physical computing tasks.
Rugged Intelligence Goes Mainstream
The prevalence of industrial-grade, wide-temperature systems like the ADL-AI2500 and various fanless computers indicates that high-performance computing is no longer confined to data centers or benign environments. We’re seeing 157 TOPS AI systems designed for -40°C to 70°C operation, which fundamentally changes where intelligence can be deployed. This enables AI in transportation, outdoor monitoring, industrial automation, and construction sites – environments where traditional computing would fail. The implications for industries like mining, agriculture, and infrastructure monitoring are enormous, as intelligence moves from cloud-based analysis to real-time edge decision-making in harsh conditions.
The Security Shift Left
Multiple winners addressed security not as an afterthought but as a foundational element. Thistle Technologies’ secure edge AI solution protecting models with hardware security controllers, Perforce’s static analysis tools catching vulnerabilities before testing, and TrustInSoft’s formal verification approach all point to a critical trend: security is moving left in the development process. As these systems control physical infrastructure and make autonomous decisions, the consequences of security failures become catastrophic. The industry is recognizing that you can’t bolt security onto intelligent physical systems – it must be designed in from the beginning using mathematically provable methods.
The Modular Future of Edge Computing
The winning entries demonstrate a clear move toward modular, scalable architectures. From Virtium’s RZ/G3E System-on-Module offering AI and non-AI variants using the same footprint to AAEON’s NDiS B562 with its DockInfinity expansion, manufacturers are embracing flexibility. This reflects the reality that edge deployments vary dramatically in requirements, and one-size-fits-all solutions don’t work. The ability to scale AI capability, I/O, or processing power while maintaining software compatibility will become increasingly important as companies deploy thousands of edge devices across diverse use cases.
Broader Industry Implications
Looking 12-24 months out, these winners suggest several major shifts. First, we’ll see increased specialization in silicon – not just CPUs versus GPUs, but processors optimized for specific physical computing tasks. Second, the line between embedded systems and enterprise computing will blur further, with technologies like Exascend’s 61.44TB SSDs bringing data center-scale storage to edge locations. Third, development tools will need to evolve dramatically to handle the complexity of heterogeneous systems combining analog, digital, AI, and security elements. The companies that can provide integrated toolchains for these complex systems will have significant competitive advantages.
The most significant long-term trend may be the democratization of advanced computing. Systems that required specialized expertise and custom hardware are becoming accessible through platforms like Avocado OS and development kits from Synaptics and others. This could accelerate innovation dramatically as more developers can create intelligent physical systems without deep hardware expertise. The real winners in this emerging landscape may be the companies that make Physical AI accessible, reliable, and secure for mainstream developers.
