7 Arduino Boards You Should Never Use for a Product
These boards all work perfectly fine for early prototypes, which is exactly what makes them dangerous, because the chip underneath simply isn’t a foundation you can carry forward into production.
So I’m going to walk you through the seven Arduino boards that you should never use as the starting point for a real product, explain why each one is a dead end, and tell you what to prototype on instead.
Board #1 – Arduino Uno R3 and Mega 2560
The Uno is the board that got many of you started, which is exactly why it keeps showing up in early product designs where it doesn’t belong.
The ATmega328P at the heart of it is an old 8-bit chip that’s underpowered, overpriced, and limited in I/O compared to modern 32-bit microcontrollers that cost less and do more.
When people run out of pins on the Uno, they usually reach for the Arduino Mega, which is built on the ATmega2560 and gives you 54 digital I/O and more memory.
Rather than choosing a chip based on the exact number of pins you need right now, always pick one that gives you room to grow, otherwise you’ll be forced to either switch microcontrollers or add an extra I/O expander chip.
But both of these chips share the same old 8-bit architecture, and neither one is something you’d want to carry forward onto a custom PCB for production, which would probably be better served by a modern 32-bit microcontroller.
Board #2 – Arduino Nano
The Nano is physically smaller so it feels product-friendly, which is why people try to embed it directly inside their enclosures.
It’s still the same ATmega328P with all the same limitations as the Uno, just in a tinier package.
And this is where I should point out that none of these Arduino boards are designed to go into a finished product.
Any FCC or CE certification they carry as dev boards doesn’t transfer to your finished product, they’re not optimized for power or cost, and an unshielded dev board with exposed headers can actually make passing emissions testing harder, not easier.
People sometimes solder Nanos onto carrier boards inside their products, stacking PCBs and adding height, cost, and reliability problems like failure under vibration and thermal cycling.
You could get away with embedding a Nano for small-scale sales testing, but if you’ve gotten to the point where you’re soldering a dev board onto another board, that’s a pretty clear sign you’re ready to move to a custom PCB.
Board #3 – Arduino Nano RP2040 Connect
The Arduino Nano RP2040 Connect looks like a strong prototyping choice because the RP2040 chip inside it is actually a solid microcontroller, dual-core with plenty of I/O and under a dollar in production quantities.
The problem isn’t the chip itself, it’s the board wrapped around it.
You’re paying around $25 for an RP2040 paired with a u-blox NINA module for Wi-Fi and Bluetooth, and that NINA module actually contains an ESP32 inside it, so you’re stuck with two microcontrollers when a single ESP32 could have handled the whole job for a fraction of the cost.
In most cases, using one microcontroller that does everything is going to be more cost-effective than paying for two separate chips, one for control and one for wireless.
Product creators often don’t realize there’s a much better alternative to prototyping on Arduino-branded dev boards, which is to use the chip manufacturer’s own boards instead.
If your product needs Wi-Fi, grab an Espressif ESP32 dev board.
If you only need Bluetooth Low Energy, a Nordic nRF52840 DK is a great option.
And if your product doesn’t need any wireless, an STM32 Nucleo board or a Raspberry Pi Pico will get you prototyping on a production-capable chip for just a few dollars.
These are cheaper than Arduino boards, they’re closer to what your production hardware will actually look like, and the code you write transfers much more cleanly when you’re ready to move to a custom PCB.
And if you really want to keep using the Arduino IDE, keep in mind it works with almost any of these other boards or chips.
Board #4 – Arduino MKR Series
The MKR line is Arduino’s attempt at IoT and product development, with Wi-Fi, LoRa, and cellular variants.
These boards use the SAMD21, a 32-bit ARM Cortex-M0+ chip from Microchip, which is a fine microcontroller but nothing special at the $30 to $50 price tag Arduino charges for these boards.
The MKR Wi-Fi board pairs the SAMD21 with the same kind of u-blox NINA module we just covered, which means you’re once again paying for two microcontrollers when one ESP32 could handle the whole job for less.
Where the MKR really falls apart is on the cellular side, because u-blox cellular modules are considerably more expensive than comparable options from Quectel or SIMCom, and cellular products usually have tight cost targets that make that kind of premium hard to justify.
And the LoRa version isn’t much better, since the MKR WAN boards use a Murata module built around a Semtech SX1276, while the newer Semtech SX126x family is more flexible, more power-efficient, and better supported for new designs.
So the MKR sits in an awkward middle ground, too expensive for any real production volume and too limited for complex products.
Board #5 – Arduino Uno R4 WiFi
The R4 WiFi is Arduino’s newest mainline board and it looks like a big upgrade over the R3.
It pairs a Renesas RA4M1 with an ESP32-S3 as a Wi-Fi and Bluetooth coprocessor, which means you’re paying for two chips when one would do the job.
The RA4M1 is a decent microcontroller, but it has a much smaller ecosystem than the STM32 or ESP32 families, with fewer reference designs, libraries, and community examples to pull from.
So you end up with a board that costs more than a bare ESP32-S3 dev board while adding a second chip with weaker long-term support.
Prototype on an Espressif ESP32-S3 dev board instead.
The Arduino IDE works fine to get started, but as you get closer to production you’ll probably want to switch to the full ESP-IDF toolchain.
And when you’re ready to ship, you can move to a pre-certified ESP32-S3 module with FCC and CE already done.
Board #6 – Arduino Portenta H7
The Portenta is Arduino’s high-end board, built on the STM32H747 and aimed at industrial applications.
These boards cost $100 to $120 or more, which is a lot of money for a development board.
The STM32H747 chip itself costs around $15 in modest quantities, which means you’re paying roughly a 7 to 8x markup just for the Arduino packaging around it.
But do you even really need the ultra-high performance of the H7, which is overkill for many products.
Arduino markets the Portenta as production-ready, but that price makes it completely impractical at any real volume.
And if your product truly needs STM32H7 performance, it’s complex enough that you shouldn’t be relying on the Arduino software framework anyway.
Grab an STM32H7 Nucleo or Discovery board from ST for a fraction of the price, and use STM32CubeIDE so your firmware is production-ready from day one.
If you really do need that level of performance, it might be worth considering whether a microprocessor like the STM32MP1 is a better fit than a high-end microcontroller that gets quite expensive.
Board #7 – Arduino Uno Q
The Uno Q is Arduino’s flashiest board yet, driven by their recent Qualcomm partnership.
It combines a quad-core Qualcomm Dragonwing QRB2210 microprocessor with an STM32U585 microcontroller, runs Linux Debian, supports camera and display connectivity, has AI acceleration, and starts around $50.
That sounds impressive on paper, but there’s a problem that has nothing to do with the board itself, and everything to do with Qualcomm.
Qualcomm has historically been one of the worst chip companies for startups and small teams to work with.
They don’t give you samples, they don’t give you datasheets, they don’t give you reference designs, and they don’t give you any meaningful support unless you’re a multi-billion dollar customer buying millions of chips.
The Dragonwing has recently shown up on Digi-Key, which is a step in the right direction, but a distributor listing doesn’t change decades of Qualcomm treating small customers like they don’t exist.
If you need Linux, display, and AI capabilities, there are proven alternatives like the STM32MP1, the Allwinner T113, or the Raspberry Pi Compute Module 5 which is built on the Broadcom BCM2712, all with established supply chains, accessible documentation, and dev boards you can get started on today.
If you don’t need Linux or AI, the Uno Q is wildly overpowered and overpriced compared to prototyping on an ESP32 or STM32 dev board.
So I just don’t see a reason to bet your product on Qualcomm when there are proven, well-supported alternatives that you can actually get help with.
