Space-Grade FPGAs

last updated 2026-05-04

Physics / mechanism

FPGAs hardened for space use radiation-tolerant logic fabrics that survive total ionizing dose (TID) and single-event effects (SEE) in LEO/GEO/deep-space environments. Two dominant hardening approaches: radiation-hardened-by-process (RHBP), where foundry implants and SOI substrates suppress charge collection, and radiation-hardened-by-design (RHBD), using triple-modular redundancy (TMR) and scrubbing in commercial-process FPGAs. Key vendors — Microchip (RTG4, 150 krad TID, 1.26M logic elements, 28nm FD-SOI derivative), Xilinx/AMD (Kintex Ultrascale XT via SEAKR integrators), Cobham/Aeroflex legacy parts — define the state of the art. Clock rates run 200–400 MHz; power budgets 2–15 W depending on utilization and SEL mitigation. Reprogrammability is the core value: payload logic updates on-orbit without hardware swap.

Competitive landscape

Competing approaches: ASICs offer better rad-hard performance per watt but require $10–50M NRE and multi-year schedules, viable only at high volume. Rad-hard microprocessors (LEON4, RAD750) handle control-plane tasks but lack parallelism for DSP/image processing. RHBD on COTS FPGAs (Xilinx Artix, Intel Cyclone) cuts cost 10×, acceptable for LEO CubeSats below 10 krad missions. Emerging: SRAM-based FPGAs with on-orbit scrubbing vs. flash-based non-volatile architectures (Microchip ProASIC3, Lattice iCE) trade reprogrammability for upset immunity.

Investment relevance

ApproachTID toleranceNRE / unit costReprogrammable
RHBP FPGA (RTG4)100–300 kradHigh / ~$50k+Yes
RHBD COTS FPGA10–30 kradLow / ~$500Yes
Space-grade ASIC300 krad+Very high / low @ volumeNo

GlobalFoundries has no current space-qualified FPGA process node, but its 22FDX FD-SOI platform is structurally adjacent — low leakage, body-bias tunability, and demonstrated SEE resilience make it a credible RHBD substrate. Deal-flow angle: fabless startups designing RHBD IP cores or rad-tolerant chiplets on 22FDX, or NewSpace payloads needing mid-tier TID performance without RTG4 pricing. Lunar Ventures’ embedded/edge compute thesis intersects here — FPGA-based on-orbit processing is the near-term compute architecture for SAR, hyperspectral, and comms payloads before rad-hard GPU alternatives mature.

Companies using

Connected ideas

Sources

Frontier (open questions)

Frontier questions