GM details centralized vehicle computing for 2028, with AI speeding development
The Cadillac Escalade IQ will debut the overhaul. GM also targets shorter development cycles, but its battery-design example is not a finished vehicle.
Listen to this story
The audio brief
Story brief
3 key pointsGM’s fourth-generation EV platform will move propulsion, chassis, perception and safety functions into a Central Compute Unit, with its first vehicle, the Cadillac Escalade IQ, scheduled for 2028. The redesign is intended to support more advanced driving features, while GM’s separate use of enterprise AI targets a shorter development cycle to compete with Chinese automakers. The announcement sets out technical plans and goals, not delivered capabilities: GM’s hands-off, eyes-off highway driving feature is also planned for 2028.
- 01
GM says the Central Compute Unit is 40 times more powerful than its existing systems and will reduce vehicle wiring by 30%.
- 02
GM aims to halve a traditional four-to-five-year development timeline; Chinese automakers reportedly develop vehicles in two to three years.
- 03
For a lithium-manganese-rich battery pack, engineers evaluated 300,000 virtual designs before building a physical prototype and reached program approval in one month.
GM’s next electric-vehicle platform will bring functions now scattered across the car into a central computer, laying the groundwork for more advanced automated driving. Announced at GM Forward in New York on October 21, the overhaul is scheduled to debut in the Cadillac Escalade IQ in 2028. GM is also using enterprise AI to shorten vehicle development.
The fourth-generation EV platform changes the electrical and computing foundation beneath the vehicle. Instead of distributing functions among numerous smaller chips, GM will consolidate much of that work in a Central Compute Unit, or CCU. The company says the unit is 40 times more powerful than its existing systems.
Its responsibilities extend well beyond the dashboard. The new architecture connects acceleration and power, steering, braking, active safety and software. The CCU will handle propulsion and chassis controls, along with perception—the processing that helps a vehicle understand its surroundings—and active safety functions.
That consolidation also changes how the car is assembled. GM says removing smaller chips around the vehicle will cut wiring by 30%, saving weight and simplifying assembly. Although the computer will arrive first in an electric Cadillac, GM says it will also work in fossil-fuel vehicles.
This is a more detailed timetable for a direction GM had already set. Last year, it outlined plans to overhaul vehicle electronics and computing for faster software, more capable automated driving and a conversational AI assistant. Wednesday’s announcement adds architecture details and a first vehicle, TechCrunch reported.
GM’s comparison with its existing systems.
GM’s projected reduction from centralizing vehicle functions.
GM also plans a highway-driving feature for 2028 that would let drivers take both their hands and eyes off the road, without needing to remain ready to take over. That is a planned capability, not something delivered by Wednesday’s announcement.
The computing redesign is part of the foundation for that ambition. GM concluded that its existing platform could not support the autonomy it wanted to deliver. Bringing perception, safety and vehicle controls onto the new architecture is intended to support more advanced driving features.
The work follows substantial software investment. GM recorded $866 million in capitalized software amortization expense in 2025, reflecting software costs spread over time rather than a single year’s cash spending.
Alongside the vehicle hardware, GM described a push to change how quickly it develops cars. The company says enterprise AI tools have accelerated work across its development operations as it seeks to halve a traditional four-to-five-year timeline.
Its concrete example concerns a lithium-manganese-rich battery pack. Engineers evaluated 300,000 virtual designs before assembling the first physical prototype, GM said. The concept moved to an approved program in one month. The example puts extensive digital exploration ahead of the first physical build.
The milestone was program approval, not completion of a production vehicle. It illustrates one engineering step moving faster; GM’s broader goal remains a shorter end-to-end development cycle.
The competitive target is clear. Chinese automakers have shortened vehicle development to two or three years, according to TechCrunch, and GM intends to match that pace. Its announcement therefore addresses two different clocks: the time required to engineer a vehicle, and the 2028 arrival of the computing platform meant to support its next driving capabilities.
Story updates
Latest developments
GM says AI could shrink vehicle development to two years by 2030
General Motors wants to move vehicle engineering away from a step-by-step process toward testing many designs at once. At its October 21, 2026, GM Forward event in New York, the company said AI-assisted engineering could shorten development to roughly two years, potentially by 2030. That is a projection, not a completed vehicle-development result.
GM’s clearest example concerns aerodynamics: testing how air moves around a vehicle. Sterling Anderson, GM’s product executive vice president, said a physical wind-tunnel test takes 60 hours to run, with scheduling adding weeks. An AI simulation can return results in as little as 60 seconds, he told the Detroit Free Press.
The models are trained on GM’s historical computational fluid dynamics data—computer calculations of how air flows. Anderson described their results as “remarkably similar” to those from wind-tunnel testing. That similarity is GM’s assessment; the comparison is between a simulated result and a physical test, not between two completed vehicle programs.
The proposed payoff is more chances to revise a design before moving forward. Anderson said engineers could iterate hundreds of times per day. Rather than waiting for one test and then adjusting a concept, teams could use the faster simulations to examine many possibilities within the same working day.
A smaller player can build a good model ― what it can’t build is a century of design decisions, crash tests, supply chain history and millions of cars on the road. We can.
Faster individual tests are only one part of GM’s proposal. Anderson also described changing the order of work. Engineers and designers would make decisions concurrently, rather than handing a vehicle through a sequence of separate stages. The company says it could test hundreds of concepts in a fraction of the traditional time.
The strategy extends beyond airflow. GM said AI helps it evaluate 300,000 battery chemistry and pack iterations before building prototypes. That places the technology at the screening stage, when engineers are considering alternatives rather than testing a finished vehicle. Jason Fischer, a GM product and engineering executive, said physical testing would be reduced by 50%, according to Autoweek.
Sources
- techcrunch.comGM is betting compute and AI will help it compete with Chinese automakers | TechCrunch
Reader comments
Newest comments first. Replies stay oldest first.