VISION

Where vehicle software is going — and where we stand.

Automotive is the deepest of our three sectors — and the clearest window into where all intelligent machines are heading. Here is that transition as we see it, grounded in what the industry is actually shipping. The same discipline carries to robots and connected devices.

BMW's Neue Klasse launched in series production on a centralized computing architecture with roughly twenty times the compute of its predecessors. Volkswagen Group's joint venture with Rivian has reference vehicles in testing ahead of a 2027 zonal-architecture series launch. Around 45% of OEMs and suppliers now rank the software-defined vehicle as their top strategic priority — while regulation turns software quality into a certifiable, fleet-wide obligation. The constraint is no longer vision. It is the engineering capacity to migrate legacy embedded software into this new architecture safely, verifiably, and on schedule.

Sources: BMW Group · IEEE Spectrum · Volkswagen Group · IoT Analytics

THE OUTLOOK

Five shifts that define the next product cycle.

Zonal architectures are in series production.

BMW's Neue Klasse iX3 shipped as the first series model on a centralized computing architecture: four high-performance "superbrain" computers, roughly twenty times the compute of the previous generation, and about 600 meters less wiring per vehicle. The Volkswagen–Rivian joint venture has frozen its E/E specification and targets its first zonal series launch in 2027.

This is not a concept phase anymore. Every OEM and Tier-1 now faces the same question: how do decades of proven, distributed ECU software make it onto zonal controllers and central compute — without losing the maturity that made it proven?

Our stance

The distributed-ECU era ends within this product cycle. We engineer the migration path: consolidating legacy ECU functions onto zonal and central architectures while preserving their safety pedigree.

Software-Defined Vehicle

Sources: BMW Group · IEEE Spectrum · Volkswagen Group · IoT Analytics

Classic and Adaptive AUTOSAR will coexist for a decade.

By AUTOSAR's own 2025 application survey, the Adaptive Platform has "crossed the chasm" into early-majority adoption — alongside, not instead of, Classic. Market analyses put AUTOSAR middleware at roughly $1.6B today, growing toward $3.4B by 2030, with Europe holding over 40% of it.

The practical consequence: mixed architectures are the reality for years to come. The hard engineering isn't in either platform alone — it's in the boundary between them: coexistence, communication, and staged migration.

Our stance

Migration and coexistence engineering is a core discipline, not a side task. We work on both platforms — and on the seam between them.

Software Development

Sources: AUTOSAR · Chalmers / Software Center · industry market analyses

Germany is building an open-source SDV base layer.

Eclipse S-CORE launched in 2025 as the industry's first open-source, safety-certifiable core stack for software-defined vehicles — driven by BMW, Mercedes-Benz, Bosch, and ETAS, under a VDA-coordinated agreement that also includes Volkswagen, Continental, Vector, and ZF. First releases shipped in late 2025.

When the base layer becomes shared infrastructure, differentiation moves up the stack — into the safety-qualified functions that make one vehicle different from another.

Our stance

We help our clients build on the shared open base and engineer the differentiating, safety-qualified functions above it — on top of the open stack, not instead of it.

Software-Defined Vehicle

Sources: Eclipse Foundation · VDA · S&P Global

Compliance becomes continuous engineering.

The deadlines are concrete: UNECE R155/R156 extends to its final approval categories in July 2026. The EU Cyber Resilience Act's reporting obligations begin in September 2026 — mandatory SBOMs, 24-hour incident reporting. China's GB 44495/44496 act as a market gate since January 2026. Meanwhile ISO 26262's third edition is being drafted for the AI-and-SDV era, and SOTIF has hardened into a de-facto type-approval requirement for ADAS.

Safety and security are no longer documents assembled before an audit. They are properties a development flow either produces continuously — or doesn't.

Our stance

TARA, secure boot chains, safety cases, and audit readiness belong inside the development flow. We build them in as engineering practice, not paperwork.

Safety & Security

Sources: UK Vehicle Certification Agency · European Commission · TÜV Rheinland · Automotive IQ

AI joins the qualified toolchain.

MathWorks now ships Simulink and Polyspace copilots with MISRA C/C++:2023 and AUTOSAR-aware analysis. ETAS offers generative-AI tooling for AUTOSAR configuration. Peer-reviewed work demonstrates LLM-assisted HIL fault analysis and automated test generation at scale.

The question is no longer whether to use AI tools in embedded development — it is how to use them inside an ASIL-rated, tool-qualified workflow under ISO 26262-8.

Our stance

We use AI tooling where it demonstrably holds up — and validate as if it might not. And we go a step further: engineering AI itself as a deliverable, validated for the edge.

Edge AI Engineering

Sources: MathWorks · ETAS · peer-reviewed research (arXiv)

WHAT WE COMMIT TO

No market-size bluster. Four commitments.

Senior engineers only

The engineers you meet are the engineers who build. No bench, no rotation.

Standards-first

ISO 26262, ASPICE, and ISO/SAE 21434 thinking built into daily engineering — practiced, not laminated.

Work-package scale

Right-sized for the 2–10-engineer packages that large providers deprioritize.

German and English delivery

Native to the market where the software-defined shift is happening.

Ready to engineer the transition?

Talk to an engineer, not a salesperson — tell us where your architecture is today and where it has to go.

Talk to an engineer