GM Puts One Central Computer at the Heart of Its Next EVs

General Motors has announced its fourth-generation EV platform, pairing novel battery chemistries and revamped electric motors with an entirely new central computing platform and electrical architecture. The announcement came on Wednesday, October 21, 2026, at the company's GM Forward event in New York.

According to TechCrunch's coverage of the event, the architecture is designed to connect acceleration and power, steering, braking, active safety features and software. GM will centralize much of its vehicle functionality on a Central Compute Unit (CCU), replacing many of the separate electronic control units (ECUs), the small embedded computers that each handle one job, found in today's cars.

The architecture is set to debut in 2028 with the Cadillac Escalade IQ. GM also plans to roll out an autonomous highway driving feature in the same year that lets drivers take their hands and eyes off the road without needing to be ready to take over.

The platform was not introduced in isolation. On October 22, 2025, GM said it would launch a centralized computing platform and next-generation electrical architecture across both electric and gas-powered vehicles. Its newsroom release from that date said the revised architecture dramatically reduces the number of vehicle modules, simplifies design, improves reliability and enables faster software updates. It described a unified computing core that consolidates dozens of ECUs and coordinates every subsystem in the vehicle. A companion release the same day stated that both the eyes-off driving system and a conversational AI assistant would run on that next-generation platform, debuting in 2028.

Software is already a line item on GM's books. The company reported capitalized software amortization expense of $866 million in 2025, according to TechCrunch. Capitalized costs are recorded as assets and expensed over several years.

On infotainment, GM has dropped Apple CarPlay for nearly all of its EVs and adopted a Google-based system. MotorTrend has reported that phone projection ends in 2028, when GM's new Google-based architecture comes online. Cadillac's support documentation describes the Google Built-In feature, which provides access to Google Assistant, Google Maps and other apps.

GM's software push has a partnership parallel. Volkswagen Group struck a $5.8 billion joint venture deal with Rivian for software-defined vehicle work, according to TechCrunch. The original announcement from November 12, 2024 said Volkswagen planned to invest up to $5.8 billion in Rivian and the joint venture by 2027, starting with an initial investment of $1 billion. Rivian's newsroom described the venture as combining both partners' strengths to create software for vehicles.

The EV backdrop is less settled. In January 2026, Reuters reported that GM said it would take a $6 billion writedown related to its EV pullback. The same report said GM's EV sales fell 43% in the fourth quarter, after the loss of the federal consumer tax credit followed record EV sales in the previous three months. In April 2026, Reuters reported, citing Crain's Detroit, that GM is indefinitely delaying its next-generation full-size electric truck program, including the GMC Sierra and Chevrolet line. GM says it has invested more than $60 billion in U.S. manufacturing since 2020.

In my view, the central compute move carries a risk that is easy to overlook in the spec sheet. Consolidating steering, braking and active safety functions onto a shared compute unit reduces module count and wiring, but it also concentrates safety-critical behavior in one software stack. Redundancy, fault isolation and the discipline of over-the-air validation become the real engineering problems, and they will be judged on the road rather than on a slide.

The broader context here is timing. Both the 2025 and 2026 announcements point to 2028 for the first production vehicles on the new architecture, and the EV retrenchment reported earlier this year shows how far plans can move in a short period. Whether GM's schedule holds will depend as much on its EV volumes as on its silicon.

The upside is concrete if the program delivers. A vehicle whose functions are coordinated by one compute core can receive software changes without the hardware dependencies that have slowed updates in older designs. Fewer modules and a simpler harness should also mean fewer failure points to diagnose. For engineers, the interesting work now lies in integration: how a single compute unit schedules safety functions, how the CCU handles updates across power and steering domains, and whether the 2028 timeline leaves room to validate a hands-off, eyes-off feature to the standard it will need.