Intel launches PhiSat-1 satellite into space, aiming to use its built-in Movidius Myriad 2 Vision Processing Unit to analyze cloud cover using AI
Darrell Etherington / TechCrunch :
Context & Ripple Effects
PhiSat-1 is the far edge case for a chip line Intel has been building outward since its acquisition of Movidius. Intel's 2016 AI chip vision leaned on Nervana for the datacenter and was left incomplete until the Movidius deal closed; the edge leg then shipped as consumer dev hardware, from the $79 USB Compute Stick with a 12-core Myriad 2 to the Myriad X VPU and the Keem Bay low-power VPU.
Putting a Myriad 2 on PhiSat-1 to classify cloud cover onboard moves that same silicon from desk-side development kits to autonomous inference in orbit, where no uplink round-trip is practical — the strongest possible demonstration of the low-power edge thesis Intel has been marketing since 2016.
First-order effects
- Intel gains in-orbit validation that the Myriad 2 can run AI vision workloads in a radiation-exposed, power-constrained environment, turning a consumer dev-kit chip into flight-proven hardware.
Second-order effects
- Flight heritage on PhiSat-1 gives Intel's Movidius line — through Keem Bay and its successors — a credible entry point into aerospace and other extreme-edge markets beyond USB sticks and developer boards.
Third-order effects
- If onboard inference proves out in orbit, satellite operators have a template for processing imagery at the source rather than downlinking raw data, and Intel's AI portfolio spans datacenter (Nervana) to the far edge (Myriad) as one architecture story.
The trend: AI inference is migrating from the datacenter to the extreme edge — down to low-power VPUs and, now, into orbit — with Intel positioning Movidius silicon as the endpoint of that chain.