Intel introduces Joule, a new chip module for developers to bring computer vision tech to cheap prototypes
Nick Statt / The Verge :
Context & Ripple Effects
Joule is the second act in Intel's module strategy: after the button-sized Curie module for wearables in 2015, Intel is now targeting a bigger form factor and a harder problem — giving cheap prototypes real computer vision. The same-day announcement of Project Euclid, a RealSense camera-and-sensor module aimed at robots, shows Intel attacking the developer market from two angles at once: general-purpose compute modules and purpose-built sensing.
First-order effects
- Developers prototyping drones, robots, and smart devices get a single board that bundles vision-capable compute instead of assembling cameras, sensors, and processors themselves.
- Project Euclid, launched the same day, gives robotics builders a complementary RealSense-based option, splitting Intel's own developer pitch between a general module and a robot-specific one.
Second-order effects
- Intel's acquisition of Movidius turns into a product pipeline: the $79 USB Compute Stick with the Myriad 2 VPU arrives within a year, establishing a sub-$100 price band for add-on vision hardware that rivals must match to stay in the developer conversation.
- Cheap modules shift early-stage device design toward off-the-shelf Intel silicon, letting startups validate products before committing to custom boards — and locking them into Intel's toolchain when they scale.
Third-order effects
- The module-to-VPU sequence ends at dedicated edge silicon: by 2019 Intel's Movidius Keem Bay VPU is positioned specifically for low-power computer vision at the edge, suggesting the prototype-module business was always a feeder for a standalone edge-AI chip line.
- If the pattern holds, computer vision becomes a commodity component bought as modules and sticks rather than engineered per-device, concentrating value in whoever owns the low-power vision processor — a contest Intel entered deliberately with this module strategy.
The trend: Intel spent the late 2010s assembling an edge computer-vision stack — wearables modules, developer boards, USB sticks, then dedicated VPUs — moving vision AI from cloud servers onto cheap local hardware.