I Called the RP2040 in 2021
Raspberry Pi announced the Pico on 21st January 2021. A week later, as the chip hit the real world, I tweeted: "Also I can't emphasise how important and influential the $4 Pi Pico will be, more still the actual RP2040 microcontroller." That was exactly five years ago today.
The board was interesting. The chip was the story.
The Call
Most of the attention went to the Pico itself: Raspberry Pi's first microcontroller board, $4, dual-core ARM Cortex-M0+, decent specs for the price. Nice product, good for hobbyists and education.
But the real play was selling the silicon. The RP2040 wasn't just a chip for Raspberry Pi boards. It was a chip anyone could buy, at volume, for under a dollar in quantity. Dual-core, programmable I/O, USB, proper documentation. A reference design anyone could build on.
That's what I was pointing at. Not the board. The chip.
Where It Ended Up
Today the RP2040 is everywhere:
Keyboards. The mechanical keyboard community adopted it immediately. Drop, Keychron, dozens of custom builds. The PIO (Programmable I/O) is perfect for scanning key matrices without CPU overhead.
Third-party boards. Adafruit, SparkFun, Pimoroni, Waveshare. Everyone makes RP2040 boards now. Different form factors, different peripherals, same silicon.
Commercial products. Not just hobbyist gear. Production hardware using a chip you can buy documentation for, at a price point that makes sense at scale.
Quadcopters. Full flight controllers built on RP2040. That one still amazes me.
The ecosystem that emerged around the RP2040 is bigger than anything Raspberry Pi could have built alone. That was the bet, and it paid off.
The Pattern
This is a recurring pattern in hardware: the component matters more than the reference design.
Arduino wasn't really about the boards. It was about the ATmega chips and the toolchain that made them accessible. The boards were the gateway. The ecosystem was the product.
ESP8266 and ESP32 followed the same pattern. Espressif made chips. Everyone else made boards and products. The chips won because they were cheap, capable, and documented.
Raspberry Pi understood this with the RP2040. They weren't just making a board for hobbyists. They were establishing a silicon platform that other people would build on.
Apple's move into embedded Swift is the same pattern recognition. The battle for embedded computing isn't about individual products. It's about the development experience and ecosystem around the silicon.
Why I Was Paying Attention
I've been following Raspberry Pi since 2012, before you could even get your hands on the Model B:
The BBC Micro was one of my gateways to computing (I think a future post is in order). Raspberry Pi explicitly positioned itself as the spiritual successor.
When you follow something for that long, you develop intuition for what matters. The chip being sold separately, at volume pricing, with full documentation. That was the tell.
The Embedded Future
Embedded computing is where the next decade of interesting technology will happen. IoT (and its security failures), edge AI, wearables, infrastructure monitoring, industrial automation, off-grid decentralised systems. The interesting stuff isn't in the cloud. It's at the edge. And the silicon supply chain underneath it all is about to get very political.
When ARM took a stake in Raspberry Pi, I noted the RISC-V implications. Then the RP2350 launched with dual-architecture support: ARM Cortex-M33 and RISC-V Hazard3 cores, switchable at boot. Raspberry Pi hedged their bet. They're not locked to ARM licensing forever.
The question is whether ARM's investment was partnership or containment. Because RISC-V is the real threat to ARM's licensing model — it's open, it's free, and China is pushing ahead aggressively with RISC-V designs while the West debates export controls. ARM's play is to keep the ecosystem locked in before the alternative matures. The geopolitics of instruction set architectures deserves its own article — and it's getting one.
And the communication layer is catching up with the silicon. Devices talking directly to each other, no cloud required — mesh networks running on cheap LoRa radios, device-to-device protocols, and increasingly, direct-to-satellite connectivity that bypasses terrestrial infrastructure entirely.
That was Starlink satellites transmitting LoRa. SpaceX had quietly acquired Swarm Technologies — a company built around tiny, low-cost satellites using the LoRaWAN protocol — and started baking LoRa beacons into their Gen2 Starlink spacecraft. A $5 ESP32 board with a cheap antenna, picking up packets directly from orbit. When a microcontroller in a field can talk to a satellite without a base station, without cellular infrastructure, without anything except a line of sight to the sky — the infrastructure assumptions change completely.
The chips that win the embedded future will be cheap, capable, well-documented, and ecosystem-friendly. The RP2040 ticked all those boxes. The RP2350 hedged on architecture. The next generation will need on-chip AI acceleration, better power management, and — I'm calling it now — RISC-V and Direct-to-Device (D2D) as the default, not the option.
But a word of caution:
I said that in 2016. The RP2040 proved me half right — the chip changed everything. But the warning still stands.
Building at the edge or working with embedded systems? I'd like to hear what you're working on.
This article is part of The Long View — spotting signals and patterns before they're obvious.