What's Happening Now
Predictive Maintenance
Jet engines have thousands of parts. Traditional maintenance follows fixed schedules — replace part X every Y flight hours, whether it needs it or not. AI changes this by predicting failure before it happens.
GE Aerospace — The Leader
GE monitors 44,000+ commercial engines in real-time using physics-based ML algorithms, achieving 60% earlier detection of maintenance needs. In 2025, GE deployed an AI blade inspection tool for CFM LEAP engines (the engine on most 737 MAX and A320neo aircraft), cutting inspection time in half while improving accuracy.
Pratt & Whitney — EngineWise
Pratt & Whitney's EngineWise platform gathers real-time sensor data — temperature, pressure, fuel consumption, vibration — and applies ML to predict maintenance needs before they become problems.
Rolls-Royce — IntelligentEngine
Rolls-Royce embeds 1,000+ sensors per engine, streaming terabytes of data per flight. Their AI predicts bearing wear, vibration anomalies, and oil degradation up to 500 flight hours in advance.
What this means for you: The mechanic of 2035 reads AI dashboards alongside torque specs. The engineer building these systems needs to understand both turbine physics and machine learning.
Autonomous Flight
Aircraft that fly themselves — no pilot on board, no remote operator. Multiple companies are approaching FAA certification.
Reliable Robotics — Autonomous Cargo
The closest to FAA certification for autonomous fixed-wing flight. Reliable Robotics flew a Cessna 208B Caravan with zero human input — continuous autopilot from taxi through landing. The FAA approved their certification plan in 2023, and they secured a $17.4M US Air Force contract in late 2025. Full FAA certification target: 2028.
eVTOL Air Taxis
| Company | Status (Early 2026) | Key Detail |
|---|---|---|
| Joby Aviation | Stage 4 of 5 in FAA cert. | 850+ test flights in 2025. Commercial service anticipated 2026. |
| Archer Aviation | Final certification stage | Holds 3 of 4 required operational certificates. |
| Wisk Aero (Boeing) | Gen 6 first flight Dec 2025 | Pursuing fully autonomous — no onboard pilot. Target: 2028–2029. |
Shield AI — Autonomous Military Drones
Shield AI builds drones that fly in GPS-denied and comms-denied environments. Their Hivemind AI uses visual odometry — the drone sees where it is instead of relying on GPS. The V-BAT has a $198M Coast Guard contract. The X-BAT, unveiled October 2025, is an autonomous jet-powered fighter-class drone capable of supersonic flight.
Merlin Labs
Building autonomous pilot software for existing aircraft. They have a $105M DoD contract to integrate their system into the C-130J cargo plane, and are going public at an $800M valuation.
Generative Design
Instead of an engineer designing a part and analyzing it, AI generates thousands of possible designs that meet engineering requirements and finds solutions humans would never conceive.
Airbus Bionic Partition
Airbus partnered with Autodesk to redesign the A320 cabin partition. AI used bio-inspired computational models to explore thousands of designs. The result: 45% lighter than the traditional partition, manufactured using additive manufacturing, meeting all certification requirements including 16G crash testing.
Fleet-wide deployment could save 465,000 metric tons of CO2 annually. And this is a cabin partition — imagine this approach applied to wing structures, engine mounts, and satellite frames.
nTop — GPU-Accelerated Optimization
The US Air Force Institute of Technology used nTop to redesign CubeSat structural elements, achieving 50% weight reduction with 20% stiffness increase.
CFD Acceleration
Computational fluid dynamics simulations — how air flows over a wing, through an engine, around a rocket — are essential to aerospace design. They're also brutally slow. A single high-fidelity simulation can take days or weeks on a supercomputer.
The AI solution: train neural networks on thousands of CFD simulations, then use the trained model to predict results in seconds. These "surrogate models" run orders of magnitude faster while respecting the laws of physics.
NVIDIA PhysicsNeMo
An open-source framework for building physics-informed AI models. It combines physics equations (Navier-Stokes, Euler) with neural network training. Published research has demonstrated airfoil optimization using Physics-Informed Neural Networks (PINNs).
PhysicsX
A startup building large physics models as surrogates for CFD and FEA workloads. They raised $155M+ in Series B at a valuation near $1 billion, backed by NVIDIA, Siemens, and Temasek.
What this means for you: CFD surrogates are the clearest example of where aerospace engineering and AI/ML skills combine. If you understand both the physics and the ML, you're working at the frontier.
Air Traffic Management
Air Space Intelligence — "Waze for Air Travel"
Their product Flyways optimizes flight routes by analyzing air traffic, weather, and airport conditions in real-time. Alaska Airlines uses Flyways on 55% of its flights, delivering 3–5% fuel savings on flights over 4 hours. In 2024, it saved 1.2 million gallons of fuel (11,958 metric tons of CO2).
NASA ATM-X
NASA is developing ML-based predictive services for Trajectory Based Operations — using AI to predict where aircraft will be and optimize traffic flow across the national airspace.
FAA Modernization
The FAA is replacing 618 radars past their lifecycle and upgrading telecommunications at 4,600 sites by 2028. AI and ML will be integrated into the next-generation air traffic control system.
Satellite Operations
AI is moving into orbit — literally.
Autonomous constellation management: Satellites now independently perform station-keeping, collision avoidance, and power management. As orbital congestion increases to tens of thousands of satellites, human-managed ground control doesn't scale.
On-orbit AI computing: Starcloud launched the first satellite with an NVIDIA H100 GPU and became the first entity to train an LLM in space (2025). Satellites are becoming autonomous AI platforms.
Anomaly detection: Platforms like SatGuard ingest telemetry streams to flag anomalies and predict failures — applying the same predictive maintenance concepts used on jet engines, but for satellites.
The Trajectory
| Timeframe | What's Likely |
|---|---|
| 2026–2027 | First commercial eVTOL passenger flights (Joby, Archer). AI route optimization spreads beyond Alaska Airlines. |
| 2028–2030 | Reliable Robotics achieves FAA certification for autonomous cargo. Autonomous military wingmen enter operational testing. |
| 2030–2035 | AI-designed aircraft components become standard. CFD surrogates replace most routine simulations. Single-pilot cargo operations under discussion. |
| 2035+ | Autonomous passenger flight is the open question. Regulatory, not technical, barriers will determine the timeline. |