Open beats locked-in for edge AI.

The lock-in problem

Pick a typical edge AI System-on-Module and you inherit three dependencies at once: a proprietary pinout that fixes your carrier-board design to one family, a vendor SDK that fixes your firmware to one toolchain, and a supply roadmap you don't control. When the vendor raises prices, hits an allocation shortage, or ends a part, your only options are a board respin or a firmware rewrite — usually both.

What an open standard changes

E1M fixes the contract between the module and your carrier board — the pinout, the mechanical envelope, and the electrical interface — and publishes it under CC BY-SA 4.0 onGitHub. Two footprints cover the range: 35 × 35 mm (312 pads) and 45 × 65 mm (496 pads). Any module that conforms to a footprint drops onto a base board built for that footprint.

• Open to a second source. Any silicon vendor or module maker can build to the published footprints, so a second source from another vendor needs no carrier redesign. Today's lineup is Alif on E1M and Renesas (optionally with DeepX) on E1M-X.
• Pin-compatible by design. Change the module under your product within a footprint family — for example E1M-X V2N to V2H or V2N+M1 — without touching the carrier board.
• One SDK across vendors. TheAlp SDKwraps each vendor SDK behind one C/C++ API across Alif, Renesas and DeepX silicon.
• 10-year supply. A writtenlifecycle commitment, not a slide.

Swap silicon in one line

Because the SDK abstracts the vendor layer, changing the module within a footprint family is a configuration edit, not an engineering project. Set the module SKU in board.yaml and the SDK rebuilds your application — Yocto on the application cores, Zephyr on the real-time companion, IPC stitched in — against the same<alp/*.h> API.

Where to go next

Read theE1M open standard, browse theE1M modulesthat implement it, or see how E1M compares toother edge AI platforms.

Frequently asked questions

What does vendor-neutral edge AI mean?

Vendor-neutral means your product is not tied to one silicon supplier. With the open E1M standard the carrier-board pinout, mechanical envelope, and electrical interface are fixed and published, and the Alp SDK exposes one C/C++ API across Alif, Renesas, and DeepX silicon — so application code written against that API carries over when you move to a different module. Within a footprint family you also keep the carrier board; today the lineup is Alif on the 35 × 35 mm E1M footprint and Renesas (optionally with DeepX) on the 45 × 65 mm E1M-X footprint, so moving between vendors is a carrier change, not a firmware rewrite.

How is E1M an open standard?

The E1M specification — pinout, two mechanical footprints, and electrical interface — is published publicly under CC BY-SA 4.0 on GitHub. Anyone can build a conforming module or carrier board. It is a standard, not a single vendor’s product line.

What is a second source for a System-on-Module?

A second source is an alternative supplier you can switch to without redesigning. Traditional SoM "pin-compatibility" only spans one vendor’s own family, because the footprint is proprietary. The E1M footprints are published, so any silicon vendor or module maker can build a module that drops onto an existing E1M carrier board with no redesign. Today each footprint carries one vendor’s silicon — Alif on E1M, Renesas (optionally with a DeepX DX-M1) on E1M-X — so a cross-vendor second source on the same footprint is what the standard enables, not something the current lineup already delivers.

How do I avoid edge AI vendor lock-in?

Choose an open hardware standard plus a portable software layer. Within an E1M footprint family, changing the module under your product is a one-line som.sku edit in board.yaml — the SDK rebuilds against the same <alp/*.h> API, with no board respin. Moving between footprints (for example from an Alif module on E1M to a Renesas module on E1M-X) needs a carrier for the other footprint, but your application code against the same API carries over.