Intel has introduced the Starfire processor, a new system-on-chip (SoC) designed to deliver artificial intelligence and high-performance computing capabilities aboard satellites and spacecraft.
The processor, developed for U.S. government and aerospace missions, combines an 18A-based CPU and neural processing unit (NPU) with an Intel 3-based Xe graphics processor in a Foveros 3D package.
The design is meant to enable spacecraft to perform AI inference directly in orbit rather than transmitting large volumes of raw data back to Earth for processing and then sending the results back into orbit.
Starfire features an eight-core hybrid CPU consisting of four performance cores and four low-power efficiency cores. It also incorporates a three-tile NPU for machine learning workloads and a four-core Xe GPU with 64 execution units to accelerate graphics and parallel computing tasks.
Intel plans to offer two variants of the processor: a 10-watt low-power model delivering up to 45 trillion operations per second (TOPS) of AI performance for power-constrained satellites, while a higher-performance 35-watt version reaches up to 75 TOPS for more demanding applications.
Space orbit is a very different environment from Earth, and requires a very different processor design. In addition to leveraging Intel’s latest process nodes, Starfire incorporates protections intended to withstand the harsh conditions of space.
The chip is designed to operate across a wide range of temperatures, from -55 degrees Celsius/-67 Fahrenheit to 125C/257F and is targeted for more than a decade of operational lifetime. Intel confirmed radiation qualification testing, including evaluations for total ionizing dose and single-event effects, has begun.
The SoC also supports LPDDR5 and DDR5 memory, offers 12 PCIe Gen4 lanes for high-speed peripheral connectivity, and integrates multiple compute engines into a single package, reducing size, weight and power requirements—critical factors for spacecraft design.
The potential market for this is minuscule compared to intel’s primary PC and data center business, but it’s not about volume. It’s about showing what Intel can do beyond the PC and data center. And it’s about supporting edge technology and even the roughest and harshest of environments.
However, there is also considerable potential upside for Intel. Starfire serves as an early showcase of its 18A manufacturing process in a highly specialized market, a market that Intel wants desperately.
The CPU and NPU are fabricated using the company’s 18A node, while the GPU utilizes Intel 3 technology. This demonstrates Intel’s ability to mix manufacturing processes in the same packaging through its Foveros advanced packaging technology.
The processor is expected to be manufactured through Intel’s U.S.-based Trusted Foundry program. Initial engineering samples are scheduled for the third quarter of 2026, with production anticipated in 2027 as Intel’s 18A manufacturing ramps.

AI Infrastructure
Tesla, SpaceX Commit $16.8 Billion to Build Massive AI Chip Factory in Texas
Aug 10, 2026



