GM Bets on AI: Halving Development Cycles, Smarter OnStar, and Eyes-Off Autonomous Driving by 2028
nashnova research
General Motors announced plans to use AI to compress vehicle development from four-to-five years to roughly two, while rolling out an AI-powered OnStar and eyes-off highway autonomy by 2028 — all aimed at closing the speed gap with Chinese automakers.
Cut the dev cycle in half — how does GM plan to match China's speed?
Traditional vehicle development takes four to five years; GM's target is roughly two. This means → if delivered, GM's model-refresh cadence would approach that of leading Chinese automakers.
Chief product officer Sterling Anderson acknowledged that Chinese automakers, backed by tightly integrated supplier ecosystems, can source common parts faster — a dimension where Western OEMs lag.
GM's answer is "virtual engineering" — feeding decades of accumulated engineering and manufacturing data into AI so that design and testing run in parallel on screens, not sequentially on physical prototypes.
In plain terms = China is fast because of its supply chain; GM wants to be fast because of its data, substituting virtual iteration for physical builds.
300,000 designs screened virtually — what does AI acceleration look like in practice?
Take the new lithium-rich manganese (LMR) battery — a next-generation chemistry with higher energy density. Engineers used AI to evaluate over 300,000 combinations of chemistry and components; the span from first battery-pack concept to formal project approval was one month.
GM is also building "digital twins" — full-scale virtual replicas of vehicles and entire factories — so engineers can test design changes without waiting for physical parts.
This means → what once required building a sample, installing it in a car, and running a test loop can now happen mostly on-screen, cutting both cost and time per iteration.
OnStar goes from call center to in-car AI — what can it do?
The 30-year-old OnStar emergency service is getting a full AI overhaul, debuting in the 2027 Cadillac Lyriq EV and rolling out to more GM models afterward.
The new system ties into onboard cameras for live viewing and crash footage; the car can auto-record and upload to the OnStar app when it senses a potential hazard event, such as aggressive honking.
Drivers can use voice commands to control climate and query vehicle features. GM stresses that OnStar cannot access connected-camera footage; recordings are time-limited and encrypted.
In 2027, GM will launch a new OnStar app that consolidates the existing OnStar Guardian app and the separate Buick, Cadillac, Chevrolet, and GMC brand apps.
What is "eyes-off" autonomy — and why not follow Tesla's path?
GM announced an "all-or-nothing" approach: no driver-monitoring-required assist systems. Instead, it will launch directly with full highway autonomy — the driver can handle email or watch video while the car drives. In plain terms = either the system takes full responsibility, or the feature stays off.
Anderson, who worked on Tesla's Autopilot, argued that monitoring-required systems like Tesla FSD encourage over-trust, making them paradoxically less safe.
The system is planned to debut in 2028 on the Cadillac Escalade IQ electric SUV, then expand to GM's best-selling pickups, large SUVs, and next-generation EVs.
What changes under the hood — and what is "Mission Mode"?
GM is introducing a new centralized electrical architecture: one powerful central computer replacing dozens of scattered electronic control units, delivering 40× more computing power while cutting wiring harness volume by 30%. This means → the car's "brain" consolidates from dozens of small modules into a single hub, making software updates and feature rollouts far simpler.
Anderson also outlined a longer-term "Mission Mode" vision: a personal vehicle that can independently complete tasks — pick up a friend, collect dry cleaning.
This reflects GM's endgame thesis: the biggest commercial value of autonomy is not "hands-free driving" but turning the car into an errand-running autonomous assistant.
How should the market read this AI strategy?
GM currently holds less than 10% of the U.S. EV market; Tesla still commands over 50%. GM promised next-gen EVs priced near combustion vehicles with longer range, but gave no launch date.
Whether the dev cycle truly shrinks to two years and whether eyes-off autonomy ships on schedule in 2028 are the two key milestones that will test this AI strategy's substance.
In plain terms = the blueprint is comprehensive, but GM needs a production timetable to prove AI-driven acceleration is more than a slide deck.
市场有风险，内容仅供研究参考，不构成投资建议。