How Physical AI Will Make The Network Part Of The Machine
By 2030, networks will evolve beyond data carriers to become active participants in how machines sense, decide and act, according to telecom experts.
The future of telecommunications will see networks transform from passive data carriers to active participants in machine decision-making, according to industry experts. This evolution, dubbed 'physical AI,' will fundamentally change how machines operate across various environments from factory floors to airfields.
Imagine an autonomous baggage tractor losing satellite navigation yet continuing to function through cellular network support, which helps track its position, detect nearby drones, and provide trusted timing. When moving beyond terrestrial coverage, satellite connectivity maintains supervisor contact while the machine shifts to a more cautious operational mode.
Infrastructure Transformation
Factories are projected to install over 700,000 industrial robots annually by 2028, with mobile machines expanding beyond factory floors to yards, ports, and airfields. Unlike generative AI that operates on screens, physical AI runs continuous loops of sensing, computing, and acting through physical bodies.
The next five years will witness significant infrastructure changes: radio sites evolving into AI sites that share processors between wireless functions and AI inference; the edge becoming a decision layer where camera and lidar feeds are fused; and the core transforming into an orchestrator that predicts service quality and proactively moves workloads.
By 2030, telecom will be judged less by how many people it connects and more by how safely machines complete useful work under changing conditions.
Network As Sensor
Perhaps the most underestimated change is that networks themselves will become sensors. The first 6G standards, scheduled for finalization in early 2029, will build on integrated sensing and communication that uses radio signals to detect and track objects like radar does.
This capability could allow machines to gain perception beyond their line of sight—a forklift being warned about a person its cameras cannot yet see. However, network-sensed data will require the same validation as onboard sensors before machines can act on it.
Satellite Integration Challenges
Satellites will evolve from backup coverage to a distinct operating tier with unique physics. Standards studies show maximum round-trip propagation times of 25.77 milliseconds for low-orbit satellites at 600 kilometers and 541.46 milliseconds for geostationary ones.
The future involves machines that adapt to their operating tier—continuing approved local work, reducing speed, deferring non-essential data, or pausing tasks requiring continuous supervision when satellite delays occur.
Critical Dependencies
Two lesser-known dependencies require board-level attention: timing and position, and cryptography. GPS signal-loss events in aviation rose 220% between 2021 and 2024, making cellular networks a crucial resilient time source.
A draft federal transition plan proposes deprecating today's public-key algorithms after 2030 and disallowing them after 2035. Machines purchased now may still be operational then, and post-quantum keys are many times larger than current ones, potentially adding costly round trips over satellite links.
Leadership Priorities
One of the biggest hurdles is integration and validation, where difficult failures often appear at system boundaries—handovers leaving data stale, edge nodes overloading, asymmetric routes, timing drift, or satellite path delays.
Organizations should write decision budgets for each autonomous task, rehearse transitions in digital twins and on-site, contract for measured outcomes like on-time decisions, require crypto agility and timing resilience in purchases, and designate one owner across operations, robotics, networking, and security.
Related
