CEO of Season Group, a vertically integrated Electronics Manufacturing Services (EMS) provider, and SG Wireless, a full-stack IoT provider.
For decades, hardware companies ran on a comfortable ritual: hire a small army of engineers, design the board, enclosure, power system and every mechanical bracket from scratch, and argue with a component rep over half a cent per unit like it’s a contract negotiation for a pro athlete. Cost got sorted out later, right before the containers left the dock.
It worked, once, because owning a unique piece of hardware, top to bottom, used to be the whole competitive advantage. These days, that instinct is a liability, and most electronics leaders haven’t fully clocked it.
You can watch this play out in the auto industry, where there’s now a name for what’s happening: “Chesla,” short for Chinese-plus-Tesla, coined to describe how Tesla and Chinese EV makers are outmaneuvering legacy automakers, according to McKinsey senior partner Choi Seung-hyuk. Tesla has already gotten vehicle platform development down to 36 months, and BYD, NIO and XPeng have pushed further, to 18 to 30 months, Choi writes; German automakers, by contrast, are still working to cut their cycle from 48 months to 36, partly by sharing parts across a model the way BYD does, with 30% to 60% commonality.
The Cheap Labor Excuse Doesn’t Hold Up
It’s tempting for Western executives to write this off as a low-cost-labor story. It isn’t. Buyers aren’t cross-shopping torque curves and component layouts. They want the seats to do something interesting, the sound system to feel like a concert hall, and the voice assistant to understand them. Nobody falls in love with a wiring harness.
That’s the uncomfortable truth for electronics OEMs too: your customers were never buying your hardware; they were buying what the hardware let them do. Hardware is infrastructure now, not identity. Plenty of legacy automakers have accepted this and started licensing EV platforms outright rather than reinventing the wheel.
The Bill Already Went Up, And It’s Not Slowing Down
None of this means hardware cost stops mattering. If anything, it matters more, since the ground underneath it has already shifted, not some future forecast. AI data center demand has pulled memory chip capacity away from consumer devices. DRAM contract prices were up 171.8% year over year as of Q3 2025, Tom’s Hardware reported, with a common 32GB DDR5 kit roughly doubling in a few months. NAND flash hasn’t been spared: Kingston has confirmed its own NAND pricing rose 246% over 2025.
That’s not a hypothetical squeeze on next year’s margins. It has already happened, to components sitting in products shipping right now, which means “design for cost” can’t wait until after the engineering is done. It has to happen at the whiteboard stage, for every subsystem, not just the board with chips on it.
Stop Building Everything As One Rigid Block
There’s a quieter risk in this custom-everything approach: fragility. A bespoke, tightly coupled hardware architecture, including mechanical housing, means one delayed component can freeze your production line or send engineers back to redesign something they were proud of six months ago. The fix isn’t limited to compute modules. It’s modularity as a design philosophy across the whole product: standardized connectors, interchangeable power supplies, swappable displays and sensors, common mounting points across a product family. If a part gets stuck in backorder, you swap in a compatible alternative and keep moving.
Widen The BOM, Then Hedge It
Modularity solves the “one part fails” problem. It doesn’t solve the “one country’s export controls just changed” problem. That requires widening where components come from, rather than defaulting to the same Western suppliers out of habit. There’s real supplier depth across Asia and Europe many BOMs never touch, and the industry is already leaning this way under what MIT Sloan defines as a “China+1” strategy: diversifying supplier geography instead of concentrating it.
The discount can be real. GigaDevice’s GD32 microcontrollers, built pin-compatible with STMicroelectronics’ widely used STM32 line, typically run about 30% cheaper than their European counterpart, OEMsecrets found, while clocking noticeably faster. Distributor LCSC puts the broader pattern at 30% to 50% lower unit cost across Asian-brand microcontrollers, in its own sourcing guide. Much of the “Western brand” premium in mature-node chips is brand and support, not silicon. That support gap is real: datasheets, forums and design-in help for many of these parts are Chinese-language and Asia-focused first, which is why a design or EMS partner (Electronics Manufacturing Services, a contract manufacturer building to your design) with an actual footprint in the region matters more than the datasheet does.
The complication is that adding components from more regions, especially China, brings geopolitical exposure a Western-only BOM doesn’t carry. Rather than avoiding it, engineer around it: qualify two BOM variants, one on Western parts, one leaning on Eastern suppliers where availability is stronger. That’s a built-in hedge if trade policy turns against one side, and doubles as a product-line strategy. The two variants can serve different regions or sit side by side in the same market at different price points, as Tesla sells the Model 3 and Model Y off shared underpinnings.
Where This Leaves You
Product teams need to spend far more time on market segmentation, user experience and feature prioritization, and far less time falling in love with a schematic. Engineering talent is better spent on firmware, software and the AI layer customers interact with, while board design, sourcing and manufacturing scale belong with trusted EMS and ODM partners (Original Design Manufacturers, which design and build a product largely on their own before it’s rebranded) across a genuinely global, dual-path supplier base.
Hardware engineering isn’t going away. It’s graduating from a bespoke, single-source science project into a modular, geographically hedged platform. The winners of the next decade will be the ones who understood customers first, built the software experience around that, and made sure their supply chain could survive a bad news cycle.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
