Halving the development cycle

General Motors used its annual GM Forward event in New York on Wednesday, 21 October, to set out how it plans to keep pace with Chinese automakers that bring new models to market far faster. The company says it will use artificial intelligence and virtual engineering to shorten its four- to five-year vehicle development cycle to as little as two years. Sterling Anderson, the company's chief product officer, described the change as a shift from slow, sequential engineering to parallel work and virtual testing. He acknowledged that GM's size makes it hard to change direction, but argued that scale also produces data, decades of engineering and manufacturing records the company can now use. Under the plan, design, engineering and manufacturing would be linked digitally from the first day, a method GM calls "digital thread virtualization".

“The faster we can iterate, the better our products become.”

Current cycle, low end4

Current cycle, high end5

Target, as little as2

Current cycle, low end

4 years

Current cycle, high end

5 years

Target, as little as

2 years

Data from the article text

Battery work as a test case

GM points to its lithium manganese-rich (LMR) battery program as evidence that the method works. Engineers evaluated more than 300,000 virtual battery designs before building a physical prototype, and the team went from first concept to formal approval in one month. The LMR cells are to power the fourth-generation EV platform, which GM says will offer 33 percent higher energy density than lithium iron phosphate cells at a comparable cost. The company also claims a 20 percent gain in efficiency per kilowatt-hour. These are figures GM itself put forward, and executives said the electric vehicles under development would cost not much more than gasoline and diesel models while going farther on a full charge. Anderson declined to say when those vehicles will be available to buy.

Software, computing and OnStar

The architecture change replaces dozens of separate control units with a central computer that links propulsion, steering, braking and active safety features to software. GM claims its new compute units can be up to 40 times more powerful than its current hardware. OnStar, the 30-year-old emergency and roadside assistance service with 12 million subscribers, is being rebuilt around an in-car assistant called OnStar AI. The first model slated to receive it is the 2027 Cadillac Lyriq, and a new OnStar app is due in 2027 to take over from the older Guardian application and the separate apps used by each brand.

Eyes-off driving and the case against supervision

GM plans to debut eyes-off highway driving on the 2028 Cadillac Escalade IQ, a system that would not require drivers to watch the road or retake control in an emergency. The company then plans to extend it to its most popular gas-powered SUVs and pickups and to its next electric vehicles. Anderson argues that supervised systems such as Tesla's FSD encourage drivers to trust them more than they should, and he presents the all-or-nothing approach as protection against complacency. Super Cruise already lets drivers take their hands off the wheel on hundreds of thousands of miles of pre-mapped highway. On the longer-term "Mission Mode" concept, vehicles would complete errands such as collecting a friend from the airport or dropping off dry cleaning.

“This industry is rife with promises that were not fulfilled. I am not keen on making them.”

What remains open

GM has not said publicly when true autonomy will arrive, and the 2028 launch covers only the first eyes-off system. Chinese cars are effectively barred from the United States by tariffs and regulations, yet GM still sells vehicles in China and faces Chinese-made cars in Canada and Mexico. Chinese rivals named in coverage include BYD, Geely and Xiaomi, which have rolled out inexpensive models with the latest technology across Asia, Africa, Europe and Latin America. The announcements give dates for 2027 and 2028, but no on-sale date for the lower-cost electric vehicles.