Most auto AI features are still not paying their own bills, even after years of investment across voice tools, driver prediction systems, connected car services, and digital shopping products. A live poll run during an SBD Automotive webinar found that only 18% of AI features are profitable for most attendees. See, automakers can build AI, no one is arguing that in the year 2026, but getting those tools to earn more than they cost is an entirely different story. SBD Automotive’s Robert Fisher said, “AI in automotive is nothing new. But making AI pay for itself is still very difficult.” SBD Automotive’s Andy Qiu said the industry is looking at the wrong problem when it talks about AI inside cars. “This is not a technology problem,” Andy said. “It’s a P&L problem.” The point here is that these capabilities of artificial intelligence are not just a one-off investment into new hardware. Unlike other hardware in the car, which once installed becomes silent, the artificial intelligence in the car does not become silent every time its functions are used. Each voice request, route planning, forecast, or connection can entail additional costs through the cloud. “Every time a user interacts with an AI feature, your cloud meter is running. That’s not capex anymore. That’s ongoing opex every day, forever,” Andy said. It raises an interesting business dilemma. In case of failure, the feature is an expense item for R&D. Yet in case of success, usage could drive up the cost of operations. Thus, the automobile manufacturer would need to prove that the technology generates sufficient revenue, loyalty, data value, subscription fees, or sales assistance. Andy noted that most of the manufacturers do not have proper cost management per each individual AI component. This could mean that they would fail to identify which