Most OEMs judge an automotive data platform by how much vehicle data it can collect. The better test is how much of that data leads to a decision someone acts on.
Cubic3 Cloud gives OEMs one place to manage global connectivity and analyse vehicle telemetry, with real-time monitoring and alerts. This post explains why data volume is a poor measure of a platform, and where AI agents belong in the stack.
Why more data doesn’t make a better automotive data platform
Connected cars send more data every year
Berg Insight estimates the average connected car used about 6.2 GB of data in 2025. It expects that to reach 10.1 GB by 2030 as automakers add more over-the-air (OTA) updates and cloud diagnostics. The firm also forecasts 583.1 million cars with factory-fitted telematics by 2030.
Every extra gigabyte costs money to move and store. It only pays for itself when it answers a question somebody in the business is asking.
A useful signal changes a decision
Think about the questions an OEM needs answered. Did a software update fail on one model in one market? Are drivers still using a connected feature a month after they paid for it?
The data that answers those questions is often small. It needs to arrive on time and carry enough context to link it to a specific vehicle and software version.
Research by Capgemini found that OEM results from vehicle data have fallen short of forecasts despite heavy investment in platforms. One reason it gives is that the signals OEMs collect are often limited to a basic set.
The leaders filter before they scale
On 21 October 2026, Volkswagen Group announced an AI data flywheel built with Microsoft, Nvidia and Voxel51. Porsche plays a major part in the work alongside other Group brands and CARIAD. Voxel51 says the fleet produces data equivalent to 11,000 average drivers a day.
For a fleet that size, the hard work is choosing which driving scenes are worth training on. Nvidia’s role in the project is preparing and evaluating that data before any model sees it.
The EU Data Act makes usefulness a legal issue
From 12 September 2026, connected products placed on the EU market, including vehicles, must be designed so users can access the data they generate. That requirement sits in Article 3(1) of the EU Data Act.
Data an OEM can’t find or structure is data it will struggle to hand over when a driver or fleet asks for it.
Where AI agents fit in an automotive data platform
An AI agent is software that can read data and act on it, sometimes without a person approving each step.
Agents are already moving into automotive. At CES 2026, a live connected vehicle demo with Hyundai showed generative AI turning large volumes of vehicle data into insights OEMs and fleets could act on.
The four layers of the stack
It helps to picture the stack in four layers. At the bottom is the vehicle, where sensors and software produce raw signals. Next is connectivity, which gets those signals off the car in every market where the vehicle is sold. Above that sits the cloud data platform, which cleans and labels signals so they mean the same thing across models and countries. AI agents work at the top, querying that platform to spot problems and recommend what to do.
An agent can only be as good as the layers below it. If signals arrive late or without the vehicle’s context, the agent will still give a confident answer. It just won’t be a reliable one.
Agents need limits as well as data
More access makes an agent more useful. It also raises the cost when something goes wrong.
At an industry cybersecurity summit on 7 October 2026, NHTSA Administrator Jonathan Morrison described an OpenAI security evaluation in which an autonomous agent escaped its test sandbox and reached production infrastructure. He also warned that machine learning tools are cutting the time and cost of finding and exploiting vehicle software flaws.
The practical step is to decide, before an agent goes live, which data it can read and which actions it can take. NHTSA’s guidance for devices that connect to vehicles follows the same rule: authenticate every connection and limit its access.
What Cubic3 Cloud does with vehicle telemetry
OEMs use Cubic3 Cloud to manage that connectivity from one place and to analyse telemetry with real-time monitoring and alerts. For a closer look at how drivers adopt and use connected services, Explore3 connected vehicle analytics builds on the same data.
The question worth asking of any automotive data platform is which signals will change a decision this quarter. Cubic3 Cloud starts from the telemetry of every vehicle it connects.
WEBINAR
How to Stop Fleet Spend Leakage & Regain Cost Control
Fleet spend leakage is eroding your margins across freight and logistics operations. Learn how to regain control of your fleet spend with this free on-demand webinar now.
Frequently asked questions
What is an automotive data platform?
An automotive data platform collects signals from connected vehicles, structures them and makes them available to the teams and tools that act on them. Berg Insight forecasts 583.1 million cars with factory-fitted telematics by 2030, so the volume these platforms handle will keep growing. Cubic3 Cloud provides this for OEMs alongside global connectivity management.
How much data does a connected car generate?
Berg Insight estimates the average connected car used about 6.2 GB of data in 2025 and expects that to reach 10.1 GB by 2030. Most of the growth comes from OTA software updates and cloud diagnostics.
Where do AI agents fit in automotive data?
AI agents sit at the top of the data stack and query the platform to find issues and recommend actions. They depend on clean, structured telemetry underneath. NHTSA warned in October 2026 that machine learning tools are lowering the cost of finding vehicle software flaws, so agents also need tight access controls.
Does the EU Data Act apply to connected cars?
Yes. From 12 September 2026, connected products placed on the EU market, including vehicles, must be designed so users can access the data they generate. OEMs need data structured well enough to share when a user asks for it.
