Nothing in a car makes it to the road without passing an incredible amount of rigorous testing. December’s Council meeting introduces startups and new tech in automotive testing, simulation, and validation – from ADAS, AV, and digital twin, to sensors, data, and cross-platform integration.
Silicon Valley, California, December 12, 2025/Meeting Recap/ On December 11, 2025, the Silicon Valley auto community gathered to address one of the most critical, yet often unseen, pillars of the mobility sector: Automotive Testing & Validation. Cars are complicated systems with many functions. And we expect them to perform competently, consistently, and safely in very widely varying environments, and in highly kinetic real-world use cases. To build such a vehicle requires great vision, experience, skill, engineering…and a boatload of trial and error, a.k.a testing. This testing is expensive, but necessary. Any new innovations that can allow it to be performed better, and more economically can yield lower costs and higher safety.
To address how the sector is managing this expense while improving safety, the Autotech Council convened a meeting hosted by Volkswagen at their Innovation and Engineering Center in Belmont, CA. The meeting drew a focused crowd of approximately 65 attendees, comprising OEMs, Tier 1 suppliers, and specialized tech vendors.
The agenda featured a blend of strategic imperatives and technical granularity. Following a welcome from our host, William Lathrop (Senior Principal Scientist, Volkswagen), the morning sessions began with a reality check regarding physical testing. In a fireside chat, Reuben Sarkar (President & CEO, American Center for Mobility) discussed field testing, highlighting how modern proving grounds provide the confidential, controlled environments necessary for final validations that simulation simply cannot replace. This was balanced by a panel discussion focused on the virtual realm. Ian O'Bryan (Regional Manager West Coast, dSPACE) and Michael DiBenigno (Senior Product Marketing Manager, Parallel Domain) explored the necessity of front-loading testing. They articulated how using simulation to perform validation earlier in the lifecycle (“shifting left”) yields faster iteration, fewer costly mistakes, and better outcomes. Our two initial sessions effectively illustrated the balance between the virtual and the physical.
A dominant theme throughout the meeting was the industry's aggressive move "leftward." The "Shift Left" methodology effectively pushes testing into the earliest phases of the software development lifecycle (SDLC). By utilizing Software-in-the-Loop (SIL) and Hardware-in-the-Loop (HIL) methodologies, engineers can validate an Electronic Control Unit (ECU) or a specific line of code in a centralized compute stack long before a physical prototype vehicle is ever assembled. This approach allows for massive regression testing - automatically re-running thousands of tests to ensure new updates haven't broken existing functionality. However, the conversation also highlighted a tension between deterministic testing, where inputs and outputs are known and fixed, and the emerging use of diffusion models for simulation (SIM). Diffusion models are beginning to allow for the generation of synthetic data that mimics the chaotic unpredictability of the real world, creating edge cases that human engineers might fail to script.
Despite the enthusiasm for virtualization, the consensus remained grounded in physics. As discussed during the sessions, while you can utilize Vehicle-in-the-Loop (VIL) and Closed-Loop simulations to accelerate development, final validation must eventually happen on test tracks. In any case, the objective is to avoid “testing” among the customer fleet. Simulation is excellent for identifying logic errors and software bugs, but it struggles to perfectly replicate the friction of tires on black ice or the interference of erratic weather on a LiDAR sensor. As was poignantly noted during the discussions, reality is an unforgiving master. If real-world performance does not match SIM testing, the model is wrong, not the world. Consequently, robust Root Cause Analysis remains a significant challenge; when a test fails in a complex, multi-modal simulation, pinpointing whether the failure was in the sensor model, the environment rendering, or the actual vehicle software is a complex hurdle that the industry is still working to clear.
From an analyst’s perspective, the importance of Automotive Testing & Validation in our current era cannot be overstated. We are currently navigating a dangerous trough in the hype cycle of autonomous and software-defined vehicles (SDVs). The complexity of vehicle code has exploded - modern cars run on over 100 million lines of code - and any "build, break, fix" hardware model is financially unsustainable and dangerously slow.
The energy in the room was palpable during the innovation review, where new players showcased solutions to these precise problems. The startup presenters did a great job communicating their innovations in a 10-minute window, and they were well-received by the Q&A leaders from Valeo, Toyota, and Bosch. The innovators & startups addressed topics such as:
Beyond some rudimentary learning, the tangible goal of the Council is to foster business relationships that reduce the friction of corporate engagement, ultimately reducing the time it takes for members and innovators to get deals done. By connecting the startups inventing better mousetraps with the OEMs who desperately need them, the community ensures that safer, more reliable products get into the hands of real customers sooner.
We would like to thank our host, Volkswagen, for their hospitality in Belmont. A special thank you to our speakers, Reuben Sarkar, Ian O'Bryan, and Michael DiBenigno, as well as our Q&A leaders for their contributions. All presentations and pitch decks are now available in the Autotech Council Member Library.