StrategyMay 12, 20267 min read

The picks-and-shovels case for autonomous mobility

In every major platform shift of the last three decades, the durable value did not accrue to the application that won the first wave of attention. It accrued to the infrastructure layer underneath. AWS, Stripe, Snowflake, Cloudflare, Twilio — every one of them quietly captured the long-cycle margin while the applications above churned. Autonomous mobility will be no different.

An Autzu autonomous mobility hub at scale — infrastructure as the picks-and-shovels layer of autonomous mobility.

The pattern is consistent

Cloud computing was won by AWS, not by Salesforce or Zoom — those rode on top. Online payments were won by Stripe, not by individual merchants. Data infrastructure was won by Snowflake, not by the BI dashboards above it. The CDN, identity, observability, and messaging layers each have their own quiet $50B+ winners. The application layer churns; the infrastructure layer compounds.

There are three structural reasons this keeps happening:

  • Infrastructure is shared across all the applications above it. Every new application increases utilisation rather than fragmenting it.
  • Switching costs compound. Replacing the data layer mid-flight is harder than replacing the front-end. The infrastructure operator earns multi-year contracts as the default option.
  • Real-world moats stack. Physical infrastructure, regulatory relationships, and integrations are slower to replicate than code. Time becomes the moat.

Autonomous mobility is structurally the same shape

Take the public AV operators — Waymo, Cruise, Aurora, Motional, Zoox. Each is a vertical, application-level bet. Each has chosen to spend tens of billions building autonomy stacks. Each is fighting on the same axes: city permits, miles-per-intervention, depot operations, energy procurement, regulatory presence. The application layer is in a four-way knife-fight.

Underneath all of them sits a layer no one has yet won — the physical and software infrastructure that lets a fleet of self-driving vehicles actually operate at scale. Hubs. Energy. Robotics. Operations software. That layer is shared, capital-intensive, regulated, and compounds. It looks structurally identical to every infrastructure win of the last three decades.

Autzu builds that layer. We do not bet on which AV stack wins. We bet that whichever stacks win will need hubs, energy management, and operations software — and that they will not want to build those themselves.

Why this is a better risk profile for capital

Three properties make the infrastructure layer a more efficient deployment of investor capital than the application layer:

  • Lower stack-bet risk. We do not need to call the winner of the AV race. We earn margin from each of them.
  • Earlier revenue. The hub layer monetises today — through rideshare EV rental — long before autonomous revenue arrives. Capital does not have to wait for autonomy to be solved.
  • Compounding margins. Each new operator partner increases hub utilisation rather than fragmenting it. Marginal revenue carries 70%+ contribution margin.

The historical analogue

In the 2010s, Uber and Lyft fought for the application layer of human-driven rideshare. The single most durable win from that decade was not Uber. It was the operational infrastructure that the rideshare boom forced into existence: payment rails, mapping, mobile telemetry, driver onboarding, EV fleet rental.

The 2020s will repeat the pattern. The AV application layer is in a multi-billion-dollar race. The infrastructure layer underneath it is wide open. Autzu is the picks-and-shovels position in that race, with revenue today and a defensible operational moat as autonomy scales.

References & further reading

  • Andreessen Horowitz — Software is eating the world (and infrastructure is eating software)a16z.com
  • Bessemer Venture Partners — Cloud 100 / The infrastructure layer thesisbvp.com
  • Sequoia Capital — Why infrastructure winssequoiacap.com
  • McKinsey — Reimagining the auto industry value chainmckinsey.com