Tiny Modules, Big Power — Why Geniatech’s New i.MX95 SOMs Are So Cool

Tiny Modules, Big Power — Why Geniatech’s New i.MX95 SOMs Are So Cool

I just came across Geniatech’s latest news, and I’ve gotta say — their new System‑on‑Modules based on the NXP i.MX95 chip are seriously exciting stuff for anyone into embedded projects or edge AI builds.

They’ve launched two models — one using the OSM 1.1 standard and another on SMARC 2.1 — and both are built for demanding markets like smart retail, automation, IoT, and even medical tech. What got me interested is how they’ve packed in performance, AI acceleration, and rugged reliability all in one neat package.

Inside, you’re getting a 6‑core Arm Cortex‑A55 cluster running up to 2 GHz, plus two additional real‑time cores (M7 and M33). That’s a nice combo for projects that need multitasking between real‑time control and heavy data crunching. The AI engine — NXP’s eIQ Neutron NPU — adds some real edge intelligence, letting these modules handle things like image recognition and predictive maintenance locally instead of relying on cloud compute.

Power draw? Only 3 to 8 watts. Not bad at all for what it’s doing. And multimedia support is just as impressive — 4K video codec, GPU, and all the modern interfaces (USB 3.0, PCIe Gen 3, dual Ethernet with TSN, etc.).

What I really like is how Geniatech split things up:

The OSM version (45 × 45 mm) is tiny and solderable, perfect for small devices that still need industrial durability.

The SMARC version (82 × 50 mm) is for when you want more ports and upgrade options down the line.

Both can handle crazy temperature ranges (–40 °C to +85 °C) and include built‑in security from NXP’s EdgeLock system.

Geniatech also promises long‑term support (they’re part of NXP’s 15‑year longevity plan), which means you can actually count on these modules sticking around — a big deal if you’re building something meant to last.

They showed these off at CES 2026, and samples are supposed to start shipping this quarter. I’m already eyeing a dev kit once they drop — this looks like a prime candidate for some serious DIY edge AI tinkering.