General Motors is pursuing a major shift toward artificial intelligence-driven vehicle development that aims to compress the typical four- to five-year design process down to just over two years, according to Chief Product Officer Sterling Anderson in an interview published last week. The automaker unveiled its "intelligence-defined vehicle" roadmap at a New York City event, centering on AI models, centralized computing, virtualized engineering, and autonomous driving capabilities. Anderson described the overarching theme simply: "The car's getting smarter," with vehicles becoming "a trusted, intelligent companion" through software that learns from data rather than relying solely on deterministic code.
The virtualization strategy allows GM to run thousands of simulated tests that previously would have required physical prototypes and six-month timelines, Anderson explained. For safety testing, the company can now "kick off thousands of tests in the virtual world that give me feedback much quicker" instead of being limited to a handful of crash-test prototypes. The automaker is using transformer-based foundation models trained on decades of internal design files and analyses, supplemented with external data to improve performance. For safety-critical functions like autonomous driving, inference runs onboard the vehicle to avoid latency and connectivity issues, while features like OnStar AI interactions can round-trip to the cloud since they aren't safety-critical. GM's new common electrical architecture and centralized compute platform will roll out across the entire portfolio, allowing software engineers to work as if the company had only two models in production rather than supporting vehicles with varying configurations.
On autonomous driving, Anderson emphasized that GM is "going to actually design an autonomous system" rather than pursuing incremental driver-assist features that create complacency risks. The company will take a road-type-specific approach for personally owned vehicles, expanding outward from there, rather than the geographically bounded, speed-limited approach used for robotaxis. Anderson rejected the vision-only sensor approach adopted by some competitors, arguing that "if you scale back to vision-only, physics suggests you'll be in trouble in certain situations" since cameras cannot see through dense fog or measure instantaneous velocity like radar. Regarding data privacy concerns stemming from past FTC actions, Anderson stated the company has implemented end-to-end encryption so that "even if we had someone stupid work for us again and do something like this, they couldn't" access raw user data. On factory automation, he dismissed humanoid robots as "form over function," noting that "this form factor evolved to run away from lions on the African savanna" rather than being optimized for production environments.
GM's aggressive pivot arrives as the automaker confronts hyperefficient Chinese EV makers expanding globally while US electric vehicle demand softens, forcing the company to balance multibillion-dollar battery plant investments against ongoing software development costs. Anderson positioned the company's new LMR (lithium manganese rich) battery chemistry as delivering range close to high-nickel packs at costs closer to LFP chemistry, allowing customers to buy fourth-generation GM EVs "for less money than a competitor's profitable model without relying on subsidies." The commonized electrical architecture reduces wiring length and vehicle mass, making cars inherently more affordable across ICE, hybrid, and EV powertrains, while virtualized development cuts prototype expenses and allows vehicle business cases to pencil out at lower retail prices. Anderson emphasized that capability advances over time without raising financial costs, comparing it to how modern devices "do far more than computers years ago for less money," with OnStar AI included in the basic package rather than paywalled. The century-old automaker faces steep risks if the complex AI systems falter or development encounters delays, potentially burning through massive capital while falling further behind software-first rivals like Tesla that were built around digital architectures from inception. For a legacy manufacturer carrying substantial organizational inertia, the transformation represents an existential gamble that assumes unproven software and virtualization tools can overcome structural advantages held by companies that never needed to retrofit twentieth-century processes into twenty-first-century platforms.