AAH #797 - How to Catch Up to China Speed

Simulation is using software models to predict how a vehicle or engineering system will behave under different conditions. Here it’s framed as a key enabler for faster development by finding issues earlier, reducing the need for physical prototypes, and supporting many test scenarios digitally.

Siemens is a major industrial technology company, and in this segment it’s presented as being deeply involved in simulation and digital-twin capabilities. The discussion positions Siemens’ software ecosystem as a foundation for accelerating engineering and development.

A digital twin is a virtual, continuously useful model of a product or system that goes beyond a simple 3D CAD drawing. In this discussion, it’s described as a multiphysics, multifunction model that can connect engineering requirements, system behavior, and even how the product is produced.

Multiphysics means modeling multiple physical effects together—like mechanical behavior plus thermal or electrical effects—rather than treating them separately. In the context of a digital twin, it helps the virtual model better match real-world system interactions.

NPI cycles are the “new product introduction” timelines—how long it takes to take a product from concept through engineering and into production. The segment claims simulation and digital twins can shorten these cycles by catching issues earlier and enabling more virtual testing.

Altair is referenced as the company Siemens acquired, described as based in Troy. The segment claims that this acquisition has been a major benefit to Siemens’ overall portfolio, particularly for EV-related development speed.

Vehicle development cycle refers to the end-to-end time required to design, validate, and prepare a vehicle for production. The segment claims the showcased work helped compress that cycle and reduce the number of physical prototypes needed.

Physical prototypes are real, built versions of a vehicle or subsystem used for testing and validation. The segment argues that with a properly set up digital twin and simulation, many issues can be found virtually, reducing reliance on physical prototypes.

AI (artificial intelligence) is discussed as a way to accelerate engineering exploration by running many iterations digitally. The segment contrasts traditional simulation workflows—where you build and analyze models over a few rounds—with AI-assisted workflows that can explore thousands of scenarios.

“Heavy metal AI” is a phrase used to describe automakers applying AI with the end goal of building real hardware—vehicles and components—rather than only producing text or software outputs. In this segment, it’s used to set up a comparison between simulation/digital-twin AI workflows and the more general LLM-style AI approach.

LLM stands for large language model, a type of AI trained to generate and understand text. The segment uses it as a contrast point: the discussion is about using simulation/digital-twin tools and AI to support engineering and physical product development, not just text generation.

A software defined vehicle is one where key vehicle functions are controlled primarily by software rather than fixed hardware. That means updates, feature changes, and even some driving/vehicle behaviors can be modified through software over time.

A bill of materials is the structured list of all parts, components, and materials needed to build a product. In vehicle development, there can be multiple bill-of-materials views (design, electronics, manufacturing) that must stay consistent so the right hardware and software end up in the right place.

An electronic bill of process is a digital, structured description of how something is manufactured—steps, operations, and required inputs—so production can be planned and executed consistently. For modern vehicles, connecting this to design and electronics helps reduce integration errors and accelerates changes.

A systems model is a unified digital representation of how vehicle subsystems interact—mechanical, electrical, electronics, and software. The goal is to coordinate requirements and behavior across domains so changes in one area don’t break assumptions elsewhere.

Reindustrialized refers to bringing or expanding manufacturing capacity back into a region, often through new investment and industrial policy. In this context, it’s tied to accelerating vehicle and component production by leveraging domestic manufacturing growth.

Brownfield refers to previously developed land that may need cleanup or infrastructure upgrades before it can be used for new manufacturing. The speaker’s point is that reusing existing industrial sites can be faster than starting from scratch, but it still requires careful planning and evolution of the existing structure.

Micro credentialing is a training approach where learners earn smaller, specific credentials tied to particular skills (rather than only a full degree). In an automotive engineering context, it’s used to ramp students up earlier on practical tools and workflows—especially as AI becomes part of the engineering process.

CAD (computer-aided design) environments are software tools used to create and modify vehicle designs and engineering geometry. When the host mentions “CAD environments” in an AI context, they’re talking about using AI to understand how designers work inside those tools and to suggest next steps.

“Edge” computing means processing data close to where it’s generated (like sensors on a factory line), rather than sending everything to a distant cloud server. That reduces latency, which helps teams make faster decisions during production.

The digital thread is the continuous flow of engineering and production data across the product lifecycle—from design through manufacturing and operations. The idea is to keep models, requirements, and results connected so changes propagate consistently.

A system model is a structured representation of how vehicle subsystems interact (for example, how software, electronics, and mechanical components behave together). It’s used to predict whether the overall system will meet requirements before real-world validation.

In this context, “agents” are AI systems that can take actions toward a goal (not just answer questions), such as running workflows, coordinating tasks, or triggering analysis. The speaker frames them as a supplement to human labor for engineering and simulation-heavy work.

AI tokens are the basic chunks of text (or other data) that an AI model processes. The speaker uses “token” cost as a way to describe why deploying AI agents can be expensive at first—because usage scales with how much the model processes.

“Fortress North America” is a policy/strategy concept for insulating the region’s auto supply chain and manufacturing from external shocks. In this segment, it’s framed as going beyond “Fortress America” to cover the broader North American market and its supplier ecosystem.

A tariff is a government tax or duty applied to imported goods. In the context of automaking, tariffs can change the cost and sourcing decisions for parts and vehicles, which is why companies may focus on both the “how to comply” and the administrative steps involved.

The supply chain is the network of companies and steps that produce and move materials and components. The value chain is the set of activities that add value—like manufacturing, engineering, and assembly—before the product reaches customers.

Stranded capital is money invested in assets (like factories or equipment) that becomes less useful or uneconomical when business conditions change. Here, it’s tied to moving plants and the cost of abandoning or underutilizing existing capacity.

“More regional” means automakers may tailor products, sourcing, and manufacturing to specific geographic markets rather than using one global platform everywhere. This can affect which technologies and components are used in each region.

“Build it once, sell it everywhere” is the idea of using a single global product strategy and platform across many markets. The speakers argue that trade rules and compliance requirements are making that approach harder, pushing toward region-specific versions.

“Critical minerals” are specific raw materials that are essential for modern technologies but have supply risks (political, geographic, or production constraints). In this segment, they’re discussed specifically in the context of batteries, where supply of materials like rare earths can limit how quickly EV and battery production scales.

“Rare earth” refers to a group of chemical elements that are not especially rare in total, but are difficult to mine and process in many regions. They’re widely used in battery and motor supply chains, so finding alternatives is a strategy to reduce dependence on constrained sourcing.

“Tier one” is an industry supply-chain label meaning a company that supplies major components directly to automakers (OEMs). Here, the discussion contrasts coordinating battery development across many tier-one players versus a pre-competitive approach that could be shared more broadly.

“OEMs” stands for Original Equipment Manufacturers—companies that build the vehicles themselves. In the segment, OEMs are paired with battery suppliers as participants in efforts to develop battery and related technology.

“Pre-competitive” describes work done before companies compete on final products—often shared research, standards, or datasets that multiple organizations can use. The segment argues that a pre-competitive base for training data can be easier to coordinate than having many separate, duplicative efforts.

“AV training data” is the labeled sensor and driving information used to train automated-driving (AV) systems. The segment notes that automakers are installing sensors and cameras partly to generate this data, and that high-quality datasets are difficult to obtain.

The segment is referencing SAE-style autonomy levels, where “L2” and “L3” describe increasingly capable driver-assistance/automation. “L2+ / L2 plus plus” is used here as a shorthand for systems beyond basic driver assist, while “L3” implies the vehicle can handle more of the driving under certain conditions.

OTA stands for over-the-air updates: software changes sent to a car wirelessly, without visiting a dealer. In modern vehicles, OTAs can deliver bug fixes and feature updates, but they still typically require validation and safety checks before the update is allowed to install.

“Self healing” refers to an approach where the vehicle detects certain problems and automatically corrects them—without waiting for a full software update cycle. In the segment, the speaker contrasts this with traditional OTAs by describing a driver-monitoring/configuration layer that can adjust behavior while the underlying source code still goes through testing and approval.

A driver monitoring layer is software that watches driver-related inputs (like attention, behavior, or system states) and then uses that information to adjust vehicle functions. In this discussion, it’s positioned as part of a “self healing” strategy that can configure things without directly rewriting the car’s core source code.

Homologation is the formal approval/testing process required before certain vehicle software changes can be deployed, especially when they affect safety-critical systems. The speaker notes that even with “self healing,” some updates still must go through homologation and safety review, which is why OTAs remain necessary.

“Customer goodwill” here means the trust and patience customers have when a brand delivers ongoing software improvements. It’s treated as a competitive advantage because it reduces friction when updates arrive later or when features depend on future compute/AI capabilities.

Tesla is highlighted here for its customer “goodwill” around receiving new software features. The implication is that Tesla’s approach to frequent updates makes customers more accepting of ongoing improvements and changes.

“AI layers” refers to stacking machine-learning systems on top of a vehicle’s underlying compute and software stack. The idea is that the AI can interpret data, learn patterns, and adjust behavior/configurations without needing a full software rewrite or a risky deep change to core code.

In this software context, “cloud” means remote servers used to store data, run analytics, and coordinate updates. Vehicles can upload telemetry, learn from large datasets, and then receive improved software or configurations back via OTAs.

ADS typically refers to “Autonomous Driving System,” the software and hardware stack that enables automated driving features. The discussion ties ADS to safety constraints, since updates that change how the car drives must be tightly controlled.

A configuration layer is a software abstraction where behavior can be adjusted through settings/parameters rather than changing the underlying source code. The speaker’s point is that safer bug fixes can happen by updating configuration instead of making changes that could violate safety or compliance rules.