AutomotiveCloudWatch
Three entrances

Where to begin

Ninety-odd research pieces is a lot. Read the lead piece first, then take the entrance that matches your decision. Each has five pieces and one line on why.

If you read one piece

Japan’s Quiet Monopoly Inside Every Robot Joint

Two Japanese suppliers own the premium tier of the reducers inside a robot’s joints, one layer beneath the companies that get the headlines.

Opens on Substack, where it was first published.

Executive decisions

For leaders deciding where to invest, what to trust and what to prepare for.

  1. The 18 Month Myth

    China did not shrink the engineering clock; the eighteen month launch cycle moved work upstream into suppliers, platforms and software.

  2. Jev Explained for Automotive and Manufacturing Leaders

    What TypeSafe’s Jev model does, where it fits in fleets, factories and robots, and where it does not belong yet.

  3. The 10-Second Rule for Deciding Where Automotive AI Should Run

    Three questions that settle whether automotive AI should run in the cloud, at the edge or in the machine.

  4. The Automotive Industry Is Preparing for the Wrong USMCA Fight

    The USMCA review reads as a trade negotiation, but the deeper exposure is the dependencies of a software-defined vehicle platform.

  5. The Automotive Patent Map Is Fracturing

    From January to April 2026 the patent landscape split along powertrain, AI architecture, software-defined vehicle strategy and geography at once.

Go deeper Strategy and investment leaders Research by decision

Architecture and engineering

For engineers and platform leaders choosing architectures, kernels and safety boundaries.

  1. The Automotive AI Stack, Explained

    The modern vehicle as five layers, from the code down to the battery chemistry, with AI at the center of each.

  2. The Software-Defined Vehicle Has a Second Architecture Nobody Draws

    The compute diagram every OEM draws is accurate but incomplete: a materials architecture, rare earths in the motors, sits beneath it.

  3. A Microkernel Will Not Save Your Cloud: QNX Hypervisor 8.0, Xen, and the Trusted Computing Base Nobody Publishes

    Whether QNX Hypervisor 8.0 can be a secure foundation for cloud servers, judged against Xen and the trusted computing base nobody publishes.

  4. Linux KVM Cannot Anchor an ASIL D Safety Case

    A guest-to-host escape in KVM shows why a hypervisor inside an ASIL D boundary cannot be fixed by patching fast, and what a more defensible architecture looks like.

  5. How QNX Works and Why It Is Difficult to Replicate

    The microkernel is visible; the assurance ecosystem around it is the barrier a competitor must recreate.

Go deeper Engineering and platform leaders How to evaluate an SDV platform

Factory deployment

For plant and operations leaders moving robots and AI from pilot to production.

  1. The Physical AI Stack, Explained

    Physical AI as five layers moving at different speeds, each with its own bottleneck, instead of one technology.

  2. When the Safety Case Moved Into Your Factory

    Humanoid robots are gaining capability fast, and the open question is who is responsible when they act on your factory floor.

  3. When Dave Retires, the Factory Forgets

    In plant after plant one experienced person knows what is written down nowhere, and that person is retiring.

  4. The Robot Economics Start at the Accepted Part

    The factory test for a robot sits in exceptions, quality acceptance and total operating cost, with RobCo’s Alfie as the prompt.

  5. The Unfiltered Robot Deployment Challenge

    Why a polished humanoid demo says little about what it costs to put hardware to work on a real plant floor.

Go deeper Manufacturing and operations leaders How to move an industrial AI pilot to deployment

Prefer to browse? Open the research archive →