Boardcon's PICOT536 SoM and EMT536 SBC Bring 2 TOPS Allwinner T536 Edge AI to Industrial Builds

A Compact Allwinner T536 Module Just Made Edge AI Easier for Industrial Designers

Boardcon released the PICOT536 system-on-module and the matching EMT536 single board computer on May 4, 2026, and the combination is one of the cleanest expressions of the Allwinner T536 edge AI processor we have seen on the market so far. The PICOT536 SoM measures 82×50mm and ships in configurations of 2-8GB LPDDR4 / LPDDR4X memory and 8-64GB eMMC storage, with optional WiFi 6 and Bluetooth 5.4 connectivity onboard. The EMT536 reference board pairs that module with a 180×120mm expansion platform that breaks out dual Gigabit Ethernet, the Allwinner T536's display interfaces, and expansion slots for NVMe storage and a 4G LTE modem.

For embedded designers building industrial HMI, machine vision, and robotics products, the PICOT536/EMT536 pair is the kind of module-plus-carrier package that compresses the path from prototype to production. Drop the SoM on the EMT536 to bring up firmware, validate algorithms against the 2 TOPS NPU, then re-host the same module on a custom carrier board for the production design. The mechanical and electrical interface stays consistent, and the long-term availability of the Allwinner T536 platform means designs can stay on the same SoC roadmap for the multi-year service horizon industrial customers expect.

Inside the Allwinner T536 Edge AI SoC

The Allwinner T536 is a heterogeneous edge AI SoC that combines a quad-core Arm Cortex-A55 cluster with XuanTie E907 and E902 RISC-V coprocessors and a 2 TOPS NPU. That mixed-ISA architecture is the design choice that matters most for the industrial use cases this board targets. The Cortex-A55 cluster handles the general-purpose Linux workload, the RISC-V coprocessors take real-time control loops and sensor preprocessing tasks off the main cores, and the NPU runs the lightweight inference workloads that make machine vision and HMI products feel responsive without requiring discrete accelerators.

Video Pipeline for Machine Vision

Video capability on the T536 is sized correctly for the target market: H.264 and H.265 decode up to 4K at 15fps, and encode up to 4K at 25fps. Those numbers are not chasing consumer-tier 4K60 throughput — they are tuned for industrial vision pipelines, where the more important constraints are deterministic latency, multi-stream capture, and reliable encode for storage or transmission to a supervisory system. For a single-board computer driving a smart camera, an inspection station, or a robotic vision module, the T536 video pipeline lands in the right neighborhood.

The PICOT536 SoM Form Factor and I/O

The 82×50mm SoM exposes the T536's I/O through a high-density edge connector to the carrier board. That separation of concerns is the structural win — the SoM stays in volume production for the long term, and the carrier board can be re-spun by the customer for each new product variant without needing to re-validate the core compute, memory, and power architecture. Optional WiFi 6 and Bluetooth 5.4 connectivity ride on the module itself, which is the configuration most industrial customers want when they need wireless connectivity certified once and reused across product lines.

Memory and Storage Configurations

The 2-8GB LPDDR4 / LPDDR4X memory range gives designers room to scale the SoM to the workload, and the 8-64GB eMMC storage range covers everything from minimal headless industrial deployments through media-rich HMI builds. The fact that storage and memory are configured at the module level rather than left to the carrier board is another industrial-friendly choice — it keeps the BOM simpler for downstream customers and reduces the surface area where supply chain volatility can disrupt a production line.

The EMT536 Development Board

The EMT536 reference platform integrates the PICOT536 module with the I/O expansion most evaluation customers want to see during the bring-up phase: dual Gigabit Ethernet, MIPI DSI and LVDS display outputs, camera inputs, audio in and out, USB 3.0 ports, and an M.2 slot for NVMe storage. There is also an M.2 Key-B socket paired with a NanoSIM card slot for 4G LTE or 5G cellular connectivity, which makes the EMT536 a credible bring-up platform for connected industrial deployments where the production design will need cellular fallback or primary connectivity.

Why This Matters for the Industrial Edge AI Market

The industrial edge AI segment has been waiting on a steady supply of well-balanced 2 TOPS class SoCs that pair with long-availability industrial-grade modules. The Boardcon PICOT536 is exactly that kind of part. It does not chase headline TOPS numbers, it does not push exotic process nodes, and it does not require liquid cooling or specialized power delivery. It pairs a competent Cortex-A55 cluster with the right amount of NPU acceleration for the inference budgets industrial workloads actually run, and it delivers the package in the form factor industrial customers can actually deploy.

The Setup For Embedded Designers

For makers and embedded engineers evaluating SoMs for the next generation of industrial HMI, machine vision, and robotics products, the PICOT536 and EMT536 are worth a serious look. The Allwinner T536 is a well-balanced edge AI SoC, the SoM form factor compresses the design cycle, and Boardcon has a track record of supporting industrial-grade modules across multi-year availability windows. Pricing has not been disclosed at publication time — that is the next watch item — but the architectural foundations of this announcement are exactly what the industrial edge AI segment needs more of in 2026.

Sources: CNX Software, May 4, 2026; Boardcon product page, May 2026.