Something shifted in the automotive world between 2023 and 2026 — and it was not just electric powertrains.
Artificial intelligence moved from pilot programmes into production lines, from R&D labs into supplier contracts. The gap between manufacturers adapting and those watching is widening fast.
That shift is as visible in motorsport as anywhere else. Formula 1 teams have been using AI-assisted aerodynamic development and real-time strategy modelling for years, and the same manufacturers competing at Le Mans and on the grid are now applying those lessons to their road car operations at scale. What started in the paddock is ending up on the production line.
The market right now: Pilots are over, deployments have begun
By 2026, the conversation has moved well beyond demonstrations. Waymo crossed 200,000 rides per week across the United States. Mercedes-Benz became the first carmaker to receive legal approval for Level 3 autonomous driving in Germany and is already testing Level 4 in Beijing. The prototypes have turned into products.
For OEMs, the implications are immediate. Launching a new model no longer ends at the factory gate — it now includes over-the-air update pipelines, AI model retraining, and cloud infrastructure that must remain live for ten or more years. That is a fundamentally different business from stamping metal and shipping units.
Autonomous driving: The core battleground
There is a significant difference between Level 2 driver assistance, now standard in most new vehicles, and Level 4 autonomous operation with no human fallback. Most commercial deployment in 2026 sits between L2+ and L3. Technical analysis covering the full architecture of AI in autonomous vehicles — from LiDAR-based systems to reinforcement learning in planning — outlines how the model layers powering perception, localisation, and decision-making in current production systems actually work.
The perception challenge alone is substantial. Training a neural network to identify a plastic bag blowing across a wet road at night, while a truck overtakes — that is not a demonstration problem. That is a training data problem, and solving it at scale is what separates credible autonomous programmes from marketing.
Waymo and others have moved aggressively into synthetic data generation using generative AI. The logic is straightforward: billions of edge-case kilometres are needed to train a safe model, and driving them all in the real world would create the exact safety risks the model is designed to prevent. For OEMs without proprietary data infrastructure, synthetic data is practically the only viable path to L3+ development at reasonable cost.
What AI is actually doing inside OEM operations
Step back from autonomous vehicles. AI is transforming automotive operations across the board, and much of it is less glamorous than self-driving taxis but arguably more impactful for near-term results.
BMW’s Landshut plant has been running AI-based predictive maintenance since 2022. By 2025, the system was flagging equipment failures an average of 72 hours before they occurred, cutting unplanned downtime by around 35%. Stellantis is pursuing similar infrastructure across its 40-plus global manufacturing plants, working toward a single data layer that dynamically optimises line throughput in real time.
In design and engineering, generative tools are changing how components are developed. GM’s use of Autodesk’s Generative Design platform produced seat brackets that are 40% lighter and 20% stronger than traditionally designed equivalents. Renault applied comparable approaches in its Ampere EV project to accelerate battery housing development. These are production tools on production programmes, not experiments.
The supplier side: Where the pressure is mounting
If you are a Tier-1 or Tier-2 supplier not actively building AI capability, the question is not whether the impact will arrive — it is when, and how significant it will be.
OEMs are restructuring supplier relationships around software and data integration. The specific pressures in 2026 include shorter design cycles as AI-assisted engineering compresses OEM timelines, leaving suppliers less validation lead time; software integration requirements that now extend to the API stack and OTA update pathway; real-time lifecycle telemetry demands from components; and cybersecurity compliance requirements under UN Regulation 155 that flow directly into supplier contracts.
The suppliers adapting fastest are treating software as a deliverable. Those struggling are treating AI as an add-on to an existing mechanical product mindset.
The compliance layer most suppliers are still ignoring
AI regulation in automotive is arriving faster than most OEMs and suppliers are prepared for. ISO 21448 addresses AI-introduced risks in driving automation. ISO/SAE 21434 covers cybersecurity engineering for connected vehicles. The EU AI Act classifies certain automotive AI systems as high-risk, requiring transparency mechanisms and documented human oversight, with enforcement beginning in 2026 and 2027.
Most OEMs have teams actively working on compliance. Many Tier-2 and Tier-3 suppliers do not. The assumption that AI compliance is the OEM’s problem and not the supplier’s is becoming an increasingly expensive position to hold.
The honest summary
AI is not arriving in the automotive industry — it is already operational. The question for OEMs and suppliers is not whether to engage. It is whether the organisational capabilities, supplier contracts, software architecture, and compliance frameworks are in place to compete when the pace accelerates further.
The companies moving fastest are not necessarily the largest. They are the ones that stopped treating AI as a project and started treating it as infrastructure.
