Most of what flies on a satellite today started life as a lab result five or ten years earlier. The systems in orbit now are the last research cycle reaching maturity, while the next one is already working its way through papers, trials and early demonstrations. Three threads of that next cycle are worth watching, and all three point back to the same underlying requirement: flexible, reconfigurable radio architecture.

How is AI Being Used in the Radio Chain?

Machine learning is being applied across the radio stack. At the physical layer, it's used for channel estimation and Doppler compensation in fast-moving satellite links and for anti-jamming and physical layer security. Further up the stack, it's used for beam hopping, dynamic spectrum sharing between satellite and terrestrial users and traffic prediction for capacity planning.

The obstacles are practical rather than theoretical. Satellites have limited onboard compute, which constrains how complex a model can run in real time. Channel state information can go stale by the time it's acted on, given propagation delay and satellite motion. Training data specific to the space environment is hard to come by and models trained on one scenario don't always generalise to another. None of this rules AI out. It just means the near-term reality is AI running alongside conventional signal processing rather than replacing it.

What’s Happening With Optical and Laser Links?

In March 2026, a gigabit-per-second laser link was demonstrated between an aircraft and a geostationary satellite, a world first. Optical links like this offer far more bandwidth than RF, but they come with a trade-off: they need much more precise pointing and tracking and they're more vulnerable to cloud cover and atmospheric turbulence.

The likely path forward is hybrid, not a wholesale switch from RF to optical. Optical links carry the bulk of high-bandwidth traffic where the link geometry and weather allow it, while RF continues to handle the connections that need resilience and easier tracking, including as a fallback when an optical link is degraded.

What Role will Quantum Communications Play?

Quantum key distribution promises encryption secured by the laws of physics rather than computational difficulty, using entanglement and the no-cloning theorem to detect any attempt at interception. It's a different security model to anything classical RF systems offer.

The obstacles are significant: atmospheric turbulence degrades the quantum signal, the timing precision required runs to sub-nanosecond levels, onboard quantum memory has a short coherence time and satellite payload budgets leave little room for the specialised optical hardware involved. Realistic near-term deployment looks like a layered model: quantum channels handling key exchange, with classical RF or optical channels still carrying the bulk data, rather than quantum communications replacing classical systems outright.

Why this all still comes back to flexible radio architecture

What connects these three threads is that none of them remove the need for software-defined, reconfigurable systems. AI-driven optimisation needs an architecture that can be updated as models improve. Hybrid optical-RF payloads need radio systems that can hand off between links as conditions change. Quantum-classical integration needs classical channels that can be reconfigured to work alongside a quantum layer as that technology matures. Each of these research threads raises the bar for flexibility rather than lowering it.

The MAC Ltd view

MAC Ltd is a research and development company at heart and keeping pace with where the field is heading, not just where it stands today, is part of how we support the programmes we work on.

Get in touch with MAC Ltd's expert team on +44 23 8076 7808 or email enquiries@macltd.com.