Every technology goes through a period where the presentations run well ahead of the products. SDV is in that period.
This page is the corrective — not to be cynical, but because knowing what has actually shipped is genuinely useful when you are estimating work or evaluating a claim.
What has demonstrably worked#
Over-the-air updates. Several manufacturers now routinely update vehicle software in the field, including safety-relevant components. This is real, shipped, and increasingly expected. It is also now a regulatory obligation, so it is not going away.
Cockpit consolidation. Cluster and infotainment sharing one processor is in volume production across many brands. The hypervisor-based approach works.
Modern infotainment platforms. Android Automotive and comparable stacks have shipped in large numbers with app ecosystems, voice assistants and genuine post-sale feature additions.
Zonal wiring. Real vehicles ship with real zonal harnesses and real weight savings.
What has been harder than expected#
Full central compute. Almost nobody has consolidated everything. Most vehicles are a modern cockpit and central ADAS bolted onto a conventional body electrical system, with a gateway between. That mixture is the normal case.
Service-oriented everything. Adoption is uneven. Plenty of "service-oriented" programmes have a service layer sitting on top of unchanged CAN signals — which is a legitimate migration step, but it is not the same thing.
Cross-manufacturer standards. VSS, uProtocol and the Eclipse projects are real and useful, and adoption is patchy. Most vehicles still run largely proprietary stacks.
Organisational change. This is consistently the biggest obstacle and the least discussed. Companies structured around hardware programmes and project funding find it genuinely difficult to sustain software teams for fifteen years.
What has gone badly#
Update quality. Several manufacturers have shipped over-the-air updates that degraded or disabled functions, occasionally requiring recovery visits. The capability to update everywhere quickly is also the capability to break everything quickly, and staged rollout discipline is not always as mature as the delivery mechanism.
Complexity underestimation. The consolidated software stack is genuinely harder than the sum of the ECUs it replaced. Several high-profile programmes have slipped substantially on software.
How to read an SDV claim#
Not cynically — usefully. Four questions:
- What shipped, in which model year, to how many vehicles? Demonstrations and pilots are not products.
- Which functions are actually on central compute? Frequently the cockpit only.
- What is the update cadence in the field? Monthly is a real capability. Annually is a workshop process with a different name.
- What still speaks CAN? The honest answer is usually "most of the body electronics", and that is fine — it just bounds what is possible.
Where this is heading#
The direction is not in doubt — the pressures from the previous topic are not reversible. What is uncertain is pace, and the realistic expectation is:
- Consolidation continues, unevenly, domain by domain rather than all at once
- Cockpit and ADAS lead, body electronics follow slowly
- Standards adoption grows but proprietary stacks persist for a long time
- Regulation drives update capability faster than commercial appetite would
- Feature-on-demand settles into genuinely-new capabilities rather than unlocking installed hardware
Next#
What all of this means for the work you will actually do.