In-House Automotive Chips: A Strategic Investor's View

Introduction#

From the perspective of a strategic investor in the automotive sector, the rush by leading automakers to design their own system-on-chip (SoC) and power modules is one of the most consequential capital allocation decisions in the industry today. The logic is seductive: software-defined vehicles demand differentiated silicon, supply security matters more than ever, and margins on advanced semiconductors are attractive. Yet the gap between strategic ambition and commercial viability is wide, and it is our job to underwrite that gap with discipline rather than enthusiasm.

This guide evaluates the in-house chip trend through three lenses: the economic scale thresholds required to justify custom silicon, the practical path from concept to production, and the long-term structural impact on traditional Tier 1 suppliers. For each, we translate the strategic narrative into the numbers and operating milestones that determine whether vertical integration creates value or destroys it.

The Economics of In-House SoC Development#

A custom automotive SoC is a fixed-cost business wrapped in a variable-cost one. Non-recurring engineering for a leading-edge node design, including architecture, verification, physical design, masks, and software enablement, commonly runs into the hundreds of millions of dollars, and mask sets alone at advanced nodes can exceed tens of millions. Add multi-year engineering teams, EDA licenses, IP licensing, and validation, and the program cost before the first revenue-generating vehicle ships can plausibly reach several hundred million dollars to well over a billion for a truly differentiated part.

The scale threshold follows directly. If a program costs roughly 500 million dollars and the silicon content premium over an off-the-shelf alternative is about 50 dollars per vehicle, the automaker needs on the order of ten million vehicles on that architecture to break even, before accounting for time value of money and execution risk. Even with a more generous 100-dollar premium, the hurdle remains in the five-million-unit range. In practice, only automakers with annual volumes above roughly two to three million vehicles, a multi-brand architecture strategy, and a willingness to spread one SoC family across many models can credibly clear this bar. Below that scale, the rational path is deep customization of merchant silicon through close partnerships rather than full in-house design.

Power Modules: A Different Scale Equation#

Power modules, particularly silicon carbide traction inverters, follow different economics. The technology is less about leading-edge lithography and more about packaging, thermal management, and device reliability. Non-recurring engineering is lower, often in the tens of millions, but capital expenditure on packaging lines and test capacity is substantial, and yield learning curves are steep. The per-vehicle value is high, frequently several hundred dollars, which lowers the volume threshold considerably.

Our rule of thumb for power modules is that an automaker needs roughly 500,000 to one million units annually of a given module family to justify captive design and assembly, and considerably more if it also intends to own wafer fabrication. Vertical integration here is most defensible when paired with a dual-sourcing strategy for die and a clear plan for second-life and recycling economics. Automakers that treat power modules as a pure cost-down exercise often discover that supplier scale and yield maturity beat captive economics until volumes are very large.

How to Structure a Credible In-House Program#

For automakers and investors evaluating such programs, we recommend a staged gate process. First, define the differentiation thesis in measurable terms: what does the custom SoC enable that merchant silicon cannot, and what is the quantified per-vehicle value of that capability? Second, model total cost of ownership across a ten-year horizon, including software maintenance, respins, and the opportunity cost of engineering talent diverted from vehicle programs. Third, secure architectural commitments across brands before tape-out, because demand fragmentation is the single most common cause of failure.

Operationally, the most successful programs we have seen adopt a hybrid model: own the architecture, safety logic, and software stack, but license IP and engage a design services partner for implementation. This preserves differentiation while sharing fixed costs. Finally, build a silicon lifecycle plan covering functional safety qualification, long-term supply for the aftermarket, and a clear path to a second-generation part, since a one-off chip program rarely earns back its investment.

The Long-Term Impact on Tier 1 Suppliers#

The rise of in-house silicon fundamentally re-prices the Tier 1 value proposition. When automakers internalize the SoC and the software that runs on it, Tier 1 suppliers lose the high-margin compute and power electronics layers that have historically anchored their profitability. What remains is increasingly system integration, thermal and mechanical engineering, and manufacturing services, all of which are more competitive and less defensible. We expect a bifurcation: Tier 1s that pivot to becoming silicon-agnostic integration partners, or that invest in their own differentiated power and sensor technologies, will retain relevance, while those dependent on reselling merchant compute will face sustained margin compression.

For investors, the implication is to underwrite Tier 1 exposure on the basis of content per vehicle in areas that resist commoditization, such as sensors, thermal systems, and safety-critical actuation, rather than on compute modules. The vertical integration trend is real, but it is selective, and the winners on both sides of the table will be those who understand precisely where scale thresholds fall and structure their capital accordingly.

Conclusion#

In-house automotive chip development is commercially viable only for a small set of high-volume automakers with disciplined architecture strategies and patient capital. SoC programs demand volumes in the millions, while power modules clear the bar at lower but still substantial scale. The strategic question is not whether vertical integration is desirable, but whether a given automaker's volume, software capability, and organizational patience can support it. For Tier 1 suppliers, the message is equally clear: defend the layers where differentiation persists, and treat compute as a partnership business rather than a fortress.