ARM-Based SOM vs Custom System on Module: Which is Better?

Introduction: The SOM vs Custom Dilemma in Embedded Design

Every embedded hardware team faces this fork in the road. You need a processor, memory, power management, and connectivity — but how you package that silicon determines everything from your first prototype to your ten-thousandth production unit. The choice between an ARM-based SOM and a fully custom design isn't just technical. It's a business decision that shapes your budget, timeline, and risk profile.

ARM-based System on Modules (SOMs) arrive pre-validated. They've been through thermal testing, signal integrity analysis, and often regulatory certification before you even open the box. Custom designs? They give you total freedom — freedom to optimize every trace, every component, every millimeter of board space. But that freedom comes with a price tag measured in engineering months and NRE dollars.

So which path wins? Honestly, it depends on your volume, your team, and your tolerance for risk. Let's break it down.

Why this decision matters for IoT product development

In the IoT world, time is everything. Miss your launch window and your product might as well not exist. An ARM-based SOM can slash months off your development cycle — we're talking 6 to 12 months saved compared to a ground-up custom design. That's not hyperbole. When you factor in schematic design, PCB layout, bring-up, and certification, custom designs simply take longer.

But there's another side. At scale, custom designs can cut BOM cost by 30-50%. For a product shipping 100,000 units a year, that difference pays for the engineering investment many times over. The trick is knowing where your product sits on that curve.

What is an ARM-Based System on Module (SOM)?

An ARM-based SOM is essentially a computer on a board. It integrates the processor (ARM Cortex-A, Cortex-M, or both), RAM, flash storage, power management ICs, and often wireless connectivity — all on a compact, standardized module. You design a carrier board that provides the specific I/O connectors, sensors, and interfaces your product needs, then plug the SOM onto it.

Popular ARM SOM ecosystems include:

• NXP i.MX series — widely used in industrial IoT and human-machine interfaces
• TI Sitara — strong for edge computing and real-time control
• STM32MP series — excellent for mixed Cortex-A and Cortex-M workloads
• Grinn SOMs — purpose-built for edge AI for IoT applications, with pre-integrated custom edge AI solutions

Typical architecture and components

A standard ARM SOM includes a multi-layer PCB with high-density interconnects, a processor with integrated GPU or NPU, LPDDR4 or DDR4 memory, eMMC or NAND flash, and a PMIC. Many also include Wi-Fi, Bluetooth, and Ethernet PHY. The module connects to your carrier board through board-to-board connectors or edge fingers — typically 200-400 pins handling everything from USB to MIPI CSI to Gigabit Ethernet.

Common use cases in IoT and edge computing

ARM SOMs dominate the middle of the IoT market. Think industrial gateways, medical devices, smart displays, and edge ai prototyping platforms. They're especially popular for embedded ai development because you can iterate on the carrier board while keeping the compute module stable. Need to upgrade from a dual-core to a quad-core processor? Swap the SOM, not the whole board.

Custom System on Module: When You Need Full Control

A custom System on Module is exactly what it sounds like — you design the module yourself. Every component, every trace, every power rail is yours to optimize. This isn't for the faint of heart. Custom SOM design requires deep expertise in high-speed layout, signal integrity, thermal management, and regulatory compliance.

But when you get it right, the payoff is enormous. Your module fits your exact mechanical envelope. Your pinout matches your carrier board perfectly. Your BOM cost is optimized for your specific volume. And you own the design — no dependency on a module vendor's roadmap or end-of-life decisions.

Customization potential and trade-offs

Going custom lets you solve problems that off-the-shelf SOMs can't touch. Need a module that fits inside a 2mm-thin wearable? Want a specific set of security features like secure enclave or encrypted storage? Require a non-standard operating temperature range? Custom is your only option.

The trade-offs are real, though. Development time stretches to 12-18 months. NRE for a custom SOM with six or more layers can easily run $100,000-$300,000. And you own the risk — if a component goes end-of-life, you're the one redesigning the board.

Typical scenarios for going custom

Custom SOMs make sense in three situations: high volume, extreme constraints, or specialized security. If you're shipping 100,000+ units annually, the unit cost savings dwarf the NRE. If your product has unique mechanical or thermal requirements — think aerospace, automotive, or medical implants — custom is often mandatory. And if you need FIPS 140-2 or Common Criteria certification, a custom design gives you control over the security architecture.

Key Comparison Criteria: ARM SOM vs Custom Design

Let's get specific. Here's how the two approaches stack up across the metrics that matter most.

Cost analysis: NRE vs unit price

This is the big one. ARM SOMs shift cost from NRE to unit price. You pay more per board but spend almost nothing on module development. Custom designs do the opposite — heavy upfront investment, lower per-unit cost. The breakeven point typically falls between 10,000 and 50,000 units, depending on module complexity and your team's efficiency.

Here's a real example. A client came to us needing a custom edge ai solutions platform for industrial inspection. They projected 5,000 units over two years. An ARM SOM from Grinn cost $85 per module. A custom design would have required $200k in NRE and delivered a $55 unit cost. The SOM saved them $175k total — and got them to market 10 months faster.

Time-to-market and certification

This is where ARM SOMs crush custom designs. A pre-certified module means your carrier board doesn't need to re-certify for radiated emissions or immunity — the module's certification covers that. You're looking at weeks for FCC/CE testing instead of months. For products with wireless connectivity, that's a massive advantage.

Custom designs require full certification from scratch. Every clock, every power rail, every RF trace needs to be tested and documented. Expect 4-6 months for FCC. CE adds another 2-3 months. And if you fail the first pass — which happens more often than teams admit — you're looking at board spins and retesting.

Flexibility, scalability, and supply chain risk

ARM SOMs win on flexibility at the system level. Need to change your display interface? Redesign the carrier board, not the module. Want to offer a low-cost and a high-performance variant? Use the same carrier with different SOMs. That's real product-line agility.

Custom designs win on component-level optimization. You can select exactly the right memory density, the right PMIC, the right oscillator. No paying for features you don't need. But you also own the supply chain risk. When that one PMIC goes on allocation — and it will — you're scrambling for alternatives.

Detailed Comparison: When to Choose Each Approach

Prototyping and low-volume production (1-10k units)

For low volumes, ARM SOMs are the obvious choice. The NRE savings alone justify the decision. But there's another factor: iteration speed. With an SOM, you can spin a carrier board in two weeks. Test it. Find issues. Spin another. That rapid cycle is gold during product development.

Grinn offers several ARM SOM platforms specifically designed for edge ai prototyping. Their modules include NPUs for iot machine learning workloads, pre-integrated sensor interfaces, and comprehensive software support. You get a proven compute core and focus your engineering on what makes your product unique.

Mid-volume production (10k-100k units)

This is the gray zone. At 10,000 units, the SOM's higher unit cost starts to sting. At 50,000 units, the custom design's NRE is fully amortized. The sweet spot often lies in a hybrid approach: use an ARM SOM on a custom carrier board that you've optimized for your specific I/O and mechanical requirements.

Many teams also consider a "custom SOM" at this volume — not a ground-up design, but a modified version of an existing module. Grinn offers this service, taking their standard SOM architecture and adapting it to your exact pinout, memory configuration, and thermal requirements. You get most of the cost benefit of custom with significantly less risk.

High-volume production (100k+ units)

At high volume, a fully custom SOM or SoM-like design is hard to beat. Your BOM drops by 30-50%. Your supply chain is under your control. Your product is optimized for your specific application. The challenge is execution — you need a design partner who understands high-volume manufacturing, thermal management, and certification at scale.

This is where working with an experienced design house like Grinn pays off. They've taken custom SOM designs through production runs of 500,000+ units. They understand component lifecycle management, second-source strategies, and the manufacturing constraints that make or break a high-volume product.

Verdict: Which Approach Wins for Your Project?

There's no universal winner. The right answer depends on your specific situation. But here's a decision framework that works for most projects.

Choose an ARM-based SOM when:

• Your projected volume is under 10,000 units
• You need to launch in less than 6 months
• Your team lacks deep high-speed PCB design expertise
• You want to iterate on product features quickly
• Certification timelines are critical to your launch plan

Choose a custom design when:

• Your projected volume exceeds 50,000 units
• You have unique mechanical, thermal, or security requirements
• You need total control over component selection
• Your product's BOM cost is the primary competitive factor
• You have an experienced hardware team or a trusted design partner

Decision framework and next steps

Start with your volume projection and your timeline. If you're under 10k units and need to ship in 12 months, don't overthink it — go with an ARM SOM from a proven vendor like Grinn. Their modules are built for embedded ai development and come with carrier board reference designs that accelerate your custom board work.

If you're at high volume or have extreme constraints, invest in a custom design. But don't go it alone. Partner with a design house that offers both SOM-based carrier board development and full custom SOM design. Grinn does both, which means they can guide you to the right approach — not just the one they happen to sell. That objectivity is rare, and it's worth seeking out.

One more thing: don't underestimate the value of getting to market first. A product that launches six months early can capture market share, build brand recognition, and generate revenue that funds the next generation. Sometimes the "right" technical choice is the one that gets you shipping fastest.

For most teams, the smart play is to start with an ARM SOM, prove your product in the market, then transition to a custom design at scale. That phased approach minimizes risk while maximizing long-term cost efficiency. And with a partner like Grinn, you can make that transition smoothly — using the same architecture, the same software stack, the same engineering team throughout.