Skip to main content
The Quantum Dispatch
Back to Home
Cover illustration for M5Stack Stamp-P4 Is a $12.95 RISC-V AI Board Smaller Than a Postage Stamp

M5Stack Stamp-P4 Is a $12.95 RISC-V AI Board Smaller Than a Postage Stamp

M5Stack's Stamp-P4 squeezes an ESP32-P4 RISC-V processor, 32MB PSRAM, MIPI camera and display interfaces, and 44 GPIOs into a 30x22mm board for just $12.95.

Alex Circuit
Alex CircuitMar 30, 20263 min read

M5Stack Drops the Stamp-P4: Maximum Capability in Minimum Footprint

M5Stack published the Stamp-P4 on March 29, 2026, and it is exactly what the name implies: a thumb-sized development board that packs enough capability to drive serious embedded AI and computer vision projects for the price of a fast food lunch. At 29.8 x 22.0 x 4.3mm and $12.95, the Stamp-P4 is one of the densest compute-per-dollar propositions in the maker hardware ecosystem right now.

The ESP32-P4: Espressif's Most Capable RISC-V Chip

The Stamp-P4 is built around Espressif's ESP32-P4 — the company's most powerful microcontroller yet. The chip runs a dual-core RISC-V processor at 360 MHz, complemented by a low-power third core capable of running at 40 MHz for background tasks. Crucially, the ESP32-P4 includes AI instruction extensions: hardware acceleration for neural network inference workloads that would otherwise saturate the main cores running standard instructions.

This matters for real projects. Edge AI inference — running a small image classification or object detection model directly on the microcontroller — is practical on the ESP32-P4 in a way that it simply was not on previous Espressif chips. The AI instruction extensions are not marketing copy; they are a meaningful architectural addition for anyone doing embedded machine learning work.

Memory is 32MB PSRAM for volatile storage and 16MB NOR flash for program storage. For an embedded application, that is substantial headroom.

Camera and Display Interfaces Built In

One of the Stamp-P4's most valuable features is its native MIPI interfaces: a 2-lane MIPI DSI connector for display output and a 2-lane MIPI CSI connector for camera input. Both are built in, not broken out via adapter headers. For computer vision projects — streaming camera input through an inference pipeline to a connected display — the Stamp-P4 has all the hardware you need on a single board slightly larger than your thumb.

I/O density is impressive: 44 GPIOs exposed via castellated holes around the board perimeter, an RMII Ethernet interface available via 1.27mm expansion headers, and USB Type-C for power and programming. Power consumption is well-characterized at approximately 30.76mA during active operation and 360 microamps in deep sleep — which makes the board viable for battery-powered deployments where a standard board's idle draw would kill runtime.

Optional Wi-Fi 6 and Bluetooth 5.4

The base Stamp-P4 does not include wireless connectivity. Wireless capability comes via the optional Stamp-AddOn C6 module at $7.00, which adds Wi-Fi 6, Bluetooth 5, and 802.15.4 support. The two-module approach keeps the base board minimal for wired and tethered applications while making wireless trivially available when needed.

Total cost for a wireless-enabled Stamp-P4 system is $19.95 — genuinely remarkable for a RISC-V board with AI instruction extensions, MIPI camera and display support, 44 GPIOs, and Wi-Fi 6.

Development Ecosystem

The Stamp-P4 supports Arduino IDE, Espressif's ESP-IDF framework, and PlatformIO, covering the standard embedded development toolchains. M5Stack's proprietary UiFlow2 visual programming environment also supports the board for lower-code prototyping workflows.

The Stamp-P4 is available from the M5Stack store and AliExpress starting today. For embedded developers looking for a capable, affordable RISC-V single-board computer platform with genuine AI acceleration and camera and display interfaces, it is a strong option at any price point — and at $12.95, it is exceptional value in the compact SBC space.

Sources: M5Stack Product Page (March 29, 2026), CNX Software (March 29, 2026)