Tesla is heading into multi-billion-dollar iceberg of its own making (electrek.co)

Tesla’s push to expand its Full Self-Driving (FSD) program internationally has exposed a growing rift between promises and hardware reality, sparking customer anger and legal action. Tesla long claimed cars built since 2016 contained “all the hardware for full self‑driving,” but admitted in January 2025 that HW3 (installed 2019–2023) cannot support unsupervised FSD and that HW4 (2024+) is required. After initially saying software updates for HW4 would lag HW3 by about six months, Tesla pivoted and prioritized HW4 updates; international rollouts in China, Australia and New Zealand have launched FSD only for HW4 vehicles. Meanwhile Tesla has listed FSD subscriptions (e.g., A$149/month) and offered confusing $5,000 “loyalty” discounts that don’t specifically remedy HW3 owners’ lost capability—prompting thousands to join an Australian class-action suit. For the AI/ML community the story is a cautionary tale about overpromising software-defined features that hinge on evolving compute and sensor stacks. Technically, the gap isn’t just software maturity but insufficient onboard compute on HW3, meaning either costly retrofits or compensation. With hundreds of thousands of customers having paid up to $15,000 for FSD, Tesla faces multi‑billion-dollar exposure, regulatory scrutiny, damaged resale values, and erosion of trust in the software‑as‑feature business model—issues that will influence how automakers and AI product teams communicate capability roadmaps going forward.