[email protected] | +1 (312) 555-0147 Mon-Fri 8:00 AM - 6:00 PM CST

The Hidden Cost of the Cheapest IoT Chip: A Procurement Manager's TCO Playbook

Posted on Friday 6th of November 2026 by Soren Mikkelsen

The Number That Looks Good Until You Add Everything Else

When I compare IoT chip quotes, the first thing I do is look at the unit price. I know I shouldn't. And yet I still do it—because that's the number that goes into the spreadsheet, the one your CFO wants to see.

Last quarter, I was comparing three SoC options for a new sensor gateway. The per-unit spread was roughly $0.80. Over 20,000 units, that's $16,000—not nothing. The cheapest option looked like a no-brainer on paper.

(Here's where I should mention that "no-brainer" has burned me before.)

Let me back up. I'm a procurement manager at a 180-person IoT device manufacturer. I've been managing our component budget—around $2.3 million annually—for six years now. I've negotiated with dozens of semiconductor vendors and reviewed more BOMs than I care to count (note to self: automate that process).

And the one pattern I keep seeing, across every product category, is the same: the unit price is the least reliable number in the entire quote.

What the Quote Doesn't Tell You

The first time this really hit me was back in 2022. We were designing a smart thermostat line and chose a cheap Wi-Fi SoC based purely on per-unit cost. Both vendors in the running were solid on paper. We picked the cheaper one and saved about $12,000 on the initial order.

The next eighteen months taught me a few things.

Development time. The documentation for the cheaper chip was sparse. The support forum had unanswered questions going back months. Our firmware team spent six extra weeks getting the basics working—compared to the three weeks we'd budgeted. At our loaded engineering rate, that's roughly $18,000 in unbudgeted labor.

Certification risk. The cheaper SoC had a pre-certified module option, but the module wasn't FCC-certified for the operating frequency we needed. We ended up doing our own certification testing: another $8,500 and six weeks of schedule slip. (I really should have asked about certification scope upfront.)

Supply stability. The vendor's lead times went from eight weeks to twenty-two weeks during the 2023 downturn. We had to expedite orders twice. Expedite fees: $4,300. The customer relationship damage from a delayed product launch: harder to quantify, but real.

Bottom line: our $12,000 "savings" disappeared somewhere around month four.

Why This Keeps Happening

I don't think this is anyone's fault, exactly. It's how the industry is structured.

Semiconductor vendors compete on the numbers that appear on a datasheet: unit price, power consumption, clock speed, memory. Those are easy to compare. They fit neatly into a comparison table.

What doesn't fit into the table:

  • The quality of the development ecosystem. Does the vendor maintain an actively updated SDK? Are there good examples and a responsive community? For instance, Espressif's ESP-IDF framework is a huge part of why we consider the ESP32 family at all—not just because the chips are low-cost and integrate Wi-Fi and Bluetooth. The framework gets regular updates (roughly every 3-4 months, based on their GitHub release history as of early 2025) and the docs are actually usable. That matters when your engineers are earning $85+ an hour loaded.
  • The real learning curve. A chip with a "familiar" instruction set isn't necessarily easy to develop for. We looked at an Arm Cortex-M option once that seemed straightforward, then discovered the vendor's HAL library had some serious gaps. The ESP32-C3—the C300 in our internal SKU system, for whatever reason—actually turned out to be less painful. Not because RISC-V is inherently simple, but because the toolchain and examples were more complete.
  • Long-term supply commitment. Does the vendor have a track record of keeping parts in production for 10+ years, or do they quietly discontinue the ones that don't sell well? This is especially important for industrial customers. According to Espressif's product longevity policy (espressif.com, accessed January 2025), the ESP32 series is covered by a 10-year commitment. That's the kind of thing that should be in every vendor conversation.
  • Support quality. When something goes wrong at 2 a.m. before a production run, what's the escalation path? A dedicated FAE, or a ticket queue? We've had customers call asking how do you reset a phone when the problem was actually their gateway's power supply. Similar confusion happens with vendor support tiers.

None of these show up in a price quote. But they all cost money—usually at the worst possible time.

To be fair, some of this is on the buyer. We should ask these questions upfront. But when you're comparing eight quotes across two weeks and trying to hit a deadline, the spreadsheet wins. I get why people go with the cheapest option—budgets are real.

The Cost of Getting It Wrong

Let me put some numbers on this. After tracking our procurement data across six years (I'm a spreadsheet person, I can't help it), here's what I found:

In projects where we chose a vendor based primarily on unit price, our average cost overrun was 27%. That's actual spending vs. initial budget for the component category. In projects where we evaluated TCO upfront, the average overrun was 9%.

That's an 18-point difference. On a $200,000 annual component budget, that's $36,000.

The hidden costs cluster into three areas:

  1. Engineering time — 41% of overruns. Mostly from unfamiliar toolchains, bad docs, and debugging requirements that the vendor didn't document.
  2. Certification and compliance — 29%. Modules that weren't pre-certified for our use case, or changing regulatory requirements that the vendor didn't flag.
  3. Supply chain disruptions — 22%. Expedite fees, alternate sourcing costs, and the occasional production line stop.

The remaining 8% was miscellaneous stuff—shipping delays, MOQ surprises, that sort of thing. (This was based on our internal procurement data from 2020-2025, tracking 47 project budgets. Take it with a grain of salt for your own context, but the pattern is consistent.)

What I Do Now Instead

I'm not going to pretend I've solved this completely. But our procurement policy has changed, and it's saved us real money.

We don't compare unit prices anymore—we compare TCO spreadsheets. I built one. It has 23 line items now, including development hours (based on past projects), certification costs, and a supply-risk score. It's not perfect (note to self: update the risk scoring model), but it's better than the alternative.

We ask "what's not included" before "what's the price." This one is simple but effective. Vendors who list everything upfront—even if the total looks higher—usually end up costing less. The ones who hit you with platform fees, toolchain licenses, or "premium support" charges after you've committed... that's a red flag.

We factor in the ecosystem, not just the chip. Honestly, this is why Espressif Inc. keeps making our shortlist. The ESP32 series isn't always the cheapest per unit, but the ESP-IDF framework, the community, and the overall module availability make the total calculation work in their favor. We're not loyal to any vendor—we're loyal to the number at the bottom of the spreadsheet.

We talk to the FAE before we talk to sales. The field application engineer will tell you the truth about the learning curve. Our Espressif rep—Jackie handles our account—has been upfront about what the C3 series can and can't do. That kind of honesty saves us from expensive surprises.

None of this is revolutionary. It's just a more honest way of comparing costs.

The One Question That Changes Everything

If you take nothing else away from this: ask any vendor what their chip will cost you in year two, not year one.

Year one is the quote. Year two is the reality. The chip you pick is just one line in the BOM. The real costs are in everything around it—the tools, the docs, the supply chain, the support.

And honestly? The vendors who are willing to talk about year two are usually the ones worth working with.

author-avatar
Soren Mikkelsen

Soren Mikkelsen is a telecommunications cabling analyst covering fiber cables, coaxial cables, twisted-pair network cable, patch cords, drop cable, ribbon cable, and cable assemblies. He uses IEC 60794 and IEC 61156 test frameworks to evaluate attenuation, bandwidth, impedance, return loss, bend radius, tensile load, crush resistance, jacket performance, and operating temperature. His work helps network designers, installers, utilities, and procurement teams balance transmission distance, pathway constraints, environmental exposure, termination labor, maintainability, and lifecycle cost.

Leave a Reply