GM Pairs AI-Driven Development With a New Battery, Central Computer and OnStar Overhaul
General Motors used its Onward event on October 21, 2026 to outline a wide-ranging set of EV, autonomous driving and AI plans, which it positioned as a response to Chinese automakers with technology-heavy product lines. Most of the announcements are plans or targets rather than completed programs.
The most ambitious claim concerns development speed. GM says it will use AI to halve vehicle development time, aiming for a two-year cycle through what it calls “digital thread virtualization,” which in practice means linking design and test data so much of the vehicle is built and validated in software before hardware exists. The company's historical cycle has run four to five years. The Verge reported the targets from the event.
The battery work offers the most concrete evidence for that approach. GM says it tested more than 300,000 virtual battery designs before building a single physical prototype, and that its battery team moved from initial concept to formal approval in one month. The chemistry it settled on is lithium manganese rich (LMR), a cathode chemistry that GM compares with lithium iron phosphate (LFP), a widely used alternative. GM claims LMR delivers 33 percent higher energy density than LFP at comparable cost, along with a 20 percent gain in efficiency per kilowatt-hour. The cells will power GM's fourth-generation EV platform and will be produced by Ultium Cells, the GM and LG Energy Solution joint venture, at its Spring Hill, Tennessee, factory.
On the electrical side, GM described drive units with 40 percent fewer internal parts than current hardware. It also announced a single liquid-cooled central computer intended to replace dozens of scattered electronic control units (ECUs, the small processors that manage individual functions) and heavy wiring harnesses. GM claims that unit delivers 40 times the computing power and requires 30 percent less physical wiring. A separate claim put forward compute units up to 40 times more powerful than the current hardware.
For engineers, the consolidation thesis is the interesting part. Fewer ECUs means fewer network boundaries to validate, a smaller fault surface, and a central platform that can carry over-the-air (OTA) software updates across more functions. Whether GM can deliver those gains without new integration risk, particularly around thermal management for the liquid-cooled unit, is a question the event did not settle.
Autonomy was the other headline. GM announced plans to bring eyes-off autonomous driving, where the driver is not required to watch the road, to its top-selling trucks and SUVs. Separately, GM stated that by the end of 2026, roughly 1 million GM vehicles on the road will have Super Cruise enabled through OnStar's connected platform. That figure is a company target for the year's end.
The OnStar overhaul was announced at GM's GM Forward event in New York on the same day, and TechCrunch reported the details. GM is adding a personal AI assistant, the OnStar AI Assistant, built in-house, with agentic AI, software that can carry out multi-step tasks on a user's behalf, at the center of the redesigned service. It is set to debut in the 2027 Cadillac Lyriq EV in the coming months, with more GM vehicles to follow. GM chief product officer Sterling Anderson described OnStar as shifting from a call center into a connected intelligence platform. A new OnStar mobile app is planned for 2027, syncing with in-car software and replacing the OnStar Guardian app, launched in 2020, along with the Buick, Cadillac, Chevrolet and GMC brand apps.
The lineage matters for anyone assessing the technical path. GM began using Google Cloud's Dialogflow conversational AI in 2022 to recognize specific spoken requests, and in 2023 extended it to common queries such as routing and navigation. In 2021 GM had already opened the Guardian app, with crash detection, emergency assistance and location sharing, to owners of compatible phones, including non-GM owners. The new assistant moves from intent recognition, which maps speech to a predefined action, toward generative and agentic behavior. One capability reported at the event illustrates the shift: connected vehicle cameras can switch on automatically when the vehicle senses an impending event, capture footage and upload it to the OnStar app.
Those capabilities arrive under closer regulatory scrutiny of connected-car data. The FTC finalized an order in 2026 that bars General Motors and OnStar from sharing certain consumer data with consumer reporting agencies and requires greater transparency and explicit consent from consumers. GM's Form 10-K filed January 28, 2025 describes its cybersecurity team's focus, when incidents occur, as responding to and containing the threat and minimizing impact.
In my view, the consent regime is the constraint that will shape the OnStar AI Assistant most directly. An agent that acts on vehicle sensor data and uploads footage needs a clear, auditable record of what the driver agreed to, and the FTC order sets that bar in writing.
The broader context here is that GM is attempting several major shifts at once: a compressed development cycle, a new cell chemistry, a consolidated electrical architecture, and an AI layer on top of a decades-old subscription service. Each carries execution risk, and the announcements mostly describe intent and schedule.
GM's newsroom states that the company has invested more than $60 billion in U.S. manufacturing since 2020. The Onward announcements add a software and silicon dimension to that footprint, and their success will be measured in shipped vehicles, verified battery performance and driver adoption rather than event demonstrations.
The more hopeful reading is that a shorter development cycle and fewer parts could bring better vehicles to market sooner, and that an assistant handling requests directly could make a connected service more useful than a phone queue ever was. Those outcomes depend on delivery dates that have yet to arrive.
