V2AI
V2AI means using the AI computer of a parked, plugged-in vehicle for approved compute workloads, scheduled around the vehicle's availability and the cost and carbon intensity of electricity, without ever compromising the vehicle's safety, security or readiness to drive.
Four services compete for the same hour. The optimiser picks the most valuable mix, in a strict order of priority.
The charging plan for the next trip is fixed first. Nothing else may touch it.
Store rooftop solar and power the home in the evening, cutting energy bills.
Earn from grid services and price differences when the system needs flexibility.
Run approved computing when power is cheap or green, as a controllable load that soaks up surplus.
Central car computers reach hundreds of trillions of operations per second. For AI inference, memory bandwidth matters as much as peak operations, so we match workloads to what each platform does well: batch inference of compact models, embeddings and data-light scientific jobs.
Workloads run in a hypervisor-isolated partition with secure boot and attestation. Driving and safety functions are never shared, and the partition stops instantly when the car is unplugged or needed.
Our design follows the rules carmakers already work under: UN R155 for cybersecurity management, UN R156 for software updates, ISO 26262 for functional safety and ISO 15118-20 for charging communication.
Compute is treated as one more flexible service of the car, next to vehicle-to-home and vehicle-to-grid. A single optimiser weighs, for every plugged-in hour, the value of storing energy, supporting the home, supporting the grid and computing, against the cost of electricity, its carbon intensity and battery wear. Because computing is a controllable load, it can absorb surplus solar and move away from peak hours instead of adding to them. The design follows current and emerging standards for smart and bidirectional charging, such as ISO 15118-20, so it stays compatible as V2X becomes mainstream.
The charging cable carries a slow power-line link that is ideal for negotiating energy (ISO 15118) but will never carry computing data. Just as bidirectional charging needs a bidirectional wallbox, V2AI needs an AI gateway at the charging point: a local buffer that caches models and data, talks to the car over a fast local link and to the outside world over broadband. Both can be one and the same box, installed once.
Distributed work needs verification. We use checkpointing, redundancy and spot checks so that interrupted or faulty results never reach a customer.