Artificial Intelligence in Automotive Market Size, Share, Growth and Forecast, 2025-2035

In-vehicle AI content per car multiplies through 2035

The artificial intelligence in automotive market was valued at USD 15.14 billion in 2025 and is projected to increase from USD 17.88 billion in 2026 to USD 79.85 billion by 2035, registering a CAGR of 18.09% during 2025–2035. Demand comes from safety mandates that make AI-based braking and distraction warning standard, from city navigate-on-autopilot systems moving into mass-market cars, and from automakers selling driving features as software. Asia Pacific accounted for 49.3% of 2025 value, the largest regional share, because China combines the largest vehicle output with the fastest rollout of urban NOA. The main commercial change is the shift from one camera chip per car toward central computers, licensed end-to-end driving software and paid upgrades. In December 2025, China's industry ministry granted its first Level 3 product approvals, opening a path to manufacturer-liable automated driving in designated pilot areas of Chongqing and Beijing.

Where does AI sit in the car, and what is left out?

The market counts the processors that run neural networks for driving and cabin functions, the perception, planning, monitoring and voice software that runs on them, and the services sold around those functions, including driving-feature subscriptions, over-the-air AI upgrades and data and simulation work. Automakers buy chips and software from suppliers or build them in-house, and drivers pay for some features after purchase.

The estimate excludes AI used in factories, design, sales and insurance, along with cameras, radar and lidar, infotainment silicon that runs no AI workload, robotaxi fare revenue and data-center training spending. Global vehicle output grows slowly, so value growth comes from AI content per car rising with each step up the autonomy ladder.

Driving chips lead as paid features and Level 3 add growth

AI hardware leads the component mix while AI services grow fastest

Based on component, the market is categorized into AI hardware, AI software and AI services. AI hardware led with 53.8% of 2025 value, because every ADAS-equipped car needs at least one vision processor and L2+ cars need a central computer. AI services grow fastest at 24.50% a year, as subscriptions, paid activations and over-the-air upgrades scale with the fitted fleet rather than new-car sales alone.

L2+ city driving leads the application mix while Level 3 and above grows fastest

Based on application, the market is categorized into basic ADAS (L1–L2), advanced L2+ and navigate-on-autopilot, L3 and above automated driving, driver and occupant monitoring, and voice and conversational assistants. Advanced L2+ systems led with 39.6% of 2025 value, driven by high compute content and fast adoption in China. L3 and above grows fastest at 37.50% a year from a small base as first approvals arrive.

Mandates, city NOA and design wins lift AI content per car

Braking and distraction mandates put an AI processor in every new car

In April 2024, NHTSA finalized FMVSS No. 127, which makes automatic emergency braking with pedestrian detection standard on nearly all U.S. light vehicles by September 2029, with stops from up to 62 mph and at least 360 lives saved a year. Article 6 of the EU General Safety Regulation 2019/2144 requires distraction and drowsiness warning, which needs an in-cabin camera and gaze model. Together, these rules set a long-term floor for AI content in North America and Europe, because every new car must carry the processor and software these functions require.

Urban navigate-on-autopilot reaches mass-market cars in China

Horizon Robotics' 2025 annual results announcement cites industry data showing advanced NOA in 42.6% of new intelligent vehicle sales in China in 2025, up from 21.6% in 2024. In the sub-RMB200,000 bracket, 65% of passenger car sales, NOA penetration rose from 5% to over 50% during the year. Horizon shipped 4.01 million Journey processors, 45% of them NOA-capable, and in January 2026 a single-chip urban NOA solution entered production, cutting cost.

Download Free Sample

Design wins lock in AI content for cars launching through 2030

Supplier order books already confirm in-car AI demand several years ahead. Horizon Robotics secured more than 110 new design wins in 2025, over 20 of them for its SuperDrive urban system, and nominations for more than 40 export models from 11 automakers. In January 2026, Mobileye announced a high-volume Surround ADAS win with a major U.S. automaker on EyeQ6 High, and the Mercedes-Benz CLA runs NVIDIA DRIVE AV software. These programs fix AI content for vehicles reaching showrooms through 2030.

Automakers turn driving features into software sales

Driving-assistance subscriptions, one-off feature purchases and over-the-air upgrades turn AI into recurring income after the car is sold, which makes AI services the fastest-growing component. Tesla's FSD Supervised is the best-known paid feature, while Chinese brands often bundle city NOA at no charge to win buyers and sell higher tiers later. Licensing also grows, as automakers that cannot fund a full stack license software from Horizon, Momenta, Mobileye or NVIDIA.

End-to-end models and LLM agents redraw car software

End-to-end and reasoning models replace hand-written driving rules

Driving stacks are moving from separate perception, prediction and planning modules to single neural networks trained on fleet video. Horizon's SuperDrive, which entered mass production in November 2025, is described as China's first mass-production one-segment end-to-end system, and it delivered more than 22,000 units in just over a month. NVIDIA opened its Alpamayo 2 Super reasoning model for commercial use in August 2026. Because larger models need more in-car compute, they also support hardware demand.

Cockpit and driving compute increasingly merge on central platforms

Automakers are cutting the number of control units by running cabin and driving AI on one central platform. In July 2026, Qualcomm was named BMW Group's lead compute silicon supplier for both digital cockpit and automated driving, after Snapdragon Ride Pilot debuted on the BMW iX3 in September 2025 with validation in 60 countries. Central compute favors suppliers with both product lines and squeezes single-function chip vendors.

Download Free Sample

Large-language-model voice agents open a cabin opportunity

Voice control is becoming a conversational agent that understands several requests in one sentence and acts on vehicle settings and navigation. Geely selected Cerence xUI for overseas models, and in April 2026, BYD and Cerence announced a multilingual assistant built on Cerence's CaLLM models, starting with the ATTO 2 DM-i, according to Cerence's announcement. Suppliers that run small models on the car and larger ones in the cloud can win this cabin slot.

Level 3 systems and robotaxi compute offer a growth opportunity

L3 and above is the fastest-growing application because approvals have started to arrive. China's first Level 3 approvals in December 2025 cover a Changan model and a BAIC Arcfox model in designated areas of Chongqing and Beijing. In August 2026, NVIDIA widened its DRIVE Hyperion robotaxi ecosystem with Foxconn, VinFast, Uber and HUMAIN. Each robotaxi carries many times a private car's AI content, so fleet deals add outsized value.

Liability, crash scrutiny and chip price wars slow rollout

Manufacturer liability holds back Level 3 launches

At Level 2 the driver is responsible; at Level 3 the manufacturer takes on part of that duty within approved conditions. Many automakers delay Level 3 for that reason, and UN Regulation No. 157 sets strict conditions for automated lane keeping. The result is slow volume for the highest-value systems and long payback on development. Automakers mitigate the risk by limiting Level 3 to defined roads and speeds, adding redundant compute and logging, and launching in pilot zones first.

Crash reporting and recalls slow feature rollouts

Every serious incident involving driver assistance is now on record in the U.S. The NHTSA Standing General Order requires reports of crashes in which Level 2 ADAS was in use within 30 seconds and someone was hurt or a vulnerable road user was struck; its third amendment took effect on June 16, 2025. Investigations and recalls triggered by these reports force software changes on large fleets and delay launches. Suppliers respond with driver-monitoring checks, conservative operating limits and faster over-the-air fixes.

Approvals remain country by country

There is no single global gate for supervised city driving. The RDW approved Tesla's FSD Supervised in April 2026 with validity only in the Netherlands, and EU-wide use requires a vote of member states, as the RDW explained. China grants Level 3 access model by model and city by city. Each market adds testing, mapping and homologation cost for automakers and suppliers, which slows rollouts. Harmonized UN rules and reusable safety cases reduce that burden over time.

Chip price wars in China compress value per car

Price pressure is heavy in China, where automakers compete on standard-fit NOA and urban systems now reach cars priced around RMB150,000. Chipmakers answer with cheaper single-chip designs, which widen fitment but lower hardware value per car, so hardware grows more slowly than software and services. Export controls on advanced processors add supply risk for Chinese automakers using U.S. chips. Suppliers protect margins by bundling software, licensing their stacks to other automakers and selling higher-tier upgrades.

China's city NOA rollout keeps Asia Pacific in the lead

Asia Pacific leads as China turns city NOA into a volume feature

Asia Pacific held 49.3% of 2025 value and is the fastest-growing region at 19.60% a year, and China drives most of that growth. Huawei, Horizon Robotics, Momenta and the in-house teams at BYD, Xpeng, Li Auto and Nio compete on standard-fit NOA, and nearly one in three passenger cars sold in China in 2025 carried an advanced NOA system. The first Level 3 approvals in December 2025 add a new tier. Japan and South Korea add premium L2+ and voice content.

North America combines chip designers, hands-free driving and a federal braking rule

North America held 25.1% of 2025 value and grows at 16.80% a year. Tesla's FSD Supervised, GM Super Cruise and Ford BlueCruise lead hands-free driving, and the region hosts the main AI chip designers, including NVIDIA, Qualcomm and Ambarella. FMVSS No. 127 sets a hard floor for AI-based braking by September 2029. Mobileye's January 2026 Surround ADAS win with a major U.S. automaker extends L2+ content into high-volume models later this decade.

Download Free Sample

Europe's Safety Regulation raises monitoring content per car

Europe held 21.4% of 2025 value and grows at 15.60% a year. The General Safety Regulation makes distraction warning, drowsiness warning and speed assistance standard, so cabin-monitoring content per car is high. Mercedes-Benz runs NVIDIA DRIVE AV software on the CLA, and BMW uses Snapdragon Ride Pilot on the iX3. EU-wide approval of supervised city driving is the next step, building on the installed driver-assistance hardware base.

Latin America and Middle East & Africa rely on imported content and early robotaxi permits

Latin America held 2.3% of 2025 value and grows at 17.50% a year, as Brazil and Mexico follow U.S. and EU safety specifications on export models and Chinese brands bring NOA-equipped cars to the region. The Middle East & Africa held 1.9% and grows at 19.00% a year, led by Gulf states with premium imports and early robotaxi permits.

Download Free Sample

Level 3 approvals and a Dutch FSD nod mark 2025–2026

Reasoning models and central compute define the next car AI

• End-to-end models use single networks trained on fleet video to replace rule-based planning, and open reasoning models such as NVIDIA Alpamayo extend them.
• Central compute lets one platform run cockpit and driving AI, cutting control units and favoring suppliers with both product lines.
• Single-chip NOA on lower-cost processors brings urban NOA to cars near RMB150,000, with suppliers targeting RMB100,000 models next.
• Driver monitoring through gaze and hand tracking becomes the condition for hands-free and supervised city driving approvals.
• Regulation such as FMVSS No. 127, the EU General Safety Regulation and Chinese Level 3 rules sets the minimum AI content per car.

Chipmakers, stack builders and cabin AI firms compete

Value is spread across three supplier groups that increasingly overlap. AI chip suppliers sell driving and cockpit processors, often with software bundled. Mobileye has the largest installed base of ADAS vision processors, NVIDIA leads high-end central compute with DRIVE Orin and Thor, Qualcomm pairs Snapdragon Ride with its cockpit position, and Horizon Robotics is the largest independent Chinese supplier, with Ambarella and Black Sesame Technologies competing in vision and domain processors. Driving-stack developers range from Tesla's in-house FSD and Huawei's partner-brand system to Momenta, which licenses its stack, and Bosch, which integrates complete systems. Cabin specialists include Cerence in voice AI and Seeing Machines in driver monitoring. Entry barriers are high in chips and full stacks, which favors a few platforms, while software licensing stays more open. No company reports in-vehicle AI sales on a comparable basis, so no share ranking is given.

Companies profiled and their roles

Value shifts from camera chips to licensed driving software

Credence Research's view is that value in the car is shifting from the camera chip to licensed driving software, central compute and paid features, and that China is setting the pace for the rest of the world. The largest new revenue lies in city NOA outside China, Level 3 systems and LLM voice agents. Suppliers that combine efficient chips, end-to-end software and driver monitoring are best positioned. The main risk is liability and approval delays, so the key signals to watch are EU votes on supervised driving and China's Level 3 pilots.

Fitment rates and AI content per vehicle drive the estimate

The report uses 2025 as the base year and forecasts to 2035, with CAGR calculated over 2025–2035. Market values come from Credence Research's market model, built from vehicle output, fitment rates of each AI function and AI content per vehicle; segment and regional splits are Credence Research estimates. Supporting evidence comes from regulations and type-approval notices, company filings and press releases, and industry penetration data cited in supplier filings.

City NOA goes global and Level 3 becomes a product by 2035

Through 2035, value will tilt from basic ADAS chips toward central computers, licensed end-to-end driving software and paid features. City NOA will spread from China to Europe and North America as single-chip designs cut cost, and Level 3 will move from pilot zones to more models as liability rules settle. Asia Pacific will remain the largest region, while LLM voice agents and driver monitoring become standard. The pace will depend on type approvals, crash scrutiny, liability rules and chip pricing in China.

Note: The depth of individual sections depends on data availability for this market. Volume, capacity, trade and other unit-based metrics are included where applicable.