Uber is Transforming Transportation. They Happen to Use Foundry.

A dive into Uber's AI revolution and how they are starting to rely on Palantir for their business

Let me start with something that most people get wrong about Uber. It is not a transportation company. Uber is a data company.

Uber has lately been developing many partnerships with data and automobile companies. We see NVIDIA, Waymo, even Lucid publicly partnering with Uber to expand their network.

NVIDIA is right now the world’s hottest company and everyone wants to do business with them. At GTC DC 2025, NVIDIA and Uber announced a partnership focused on the following points; scaling Uber's autonomous vehicle fleet using DRIVE AGX Hyperion 10, building a joint AI data factory with a fleet of robotaxi data on the NVIDIA Cosmos platform, and enabling OTA model updates as Blackwell architecture matures. And you may think: How does Palantir fit in all of this?

Think about what happens every single second Uber operates. Thousands of GPS signals from drivers. Thousands of users opening the app. The algorithm calculating surge pricing in real time, assigning the nearest driver, predicting ETAs, running fraud detection and making sure the whole thing complies with local regulations that change depending on which city, which country... All of that, simultaneously, across 70+ countries, 150 million active monthly users, and a fleet of cars that the company doesn’t even own. It’s a big operation.

That’s without taking into account Uber’s actions regarding trains, air travel and other means of transportation other than cars. Think of Uber Eats, think of the ownership of a system of transportation without having to manufacture any cars or hardware. It’s a big infrastructure plan that needs of regulation and broad industry adoption. They happen to be using Foundry.

A Foundry link went active a couple of weeks ago, signaling that Uber is beginning to rely on the Foundry platform to perform some of their business. We will see if the link remains active and I’ll be monitoring its activity for my subscribers, as I do with over 2,500 Foundry links.

Let’s focus on fleet intelligence. Uber does not just manage human drivers — it is actively onboarding autonomous vehicle deployments with Waymo in Phoenix and Atlanta, with Waabi and with Aurora on freight. Each of these partnerships runs its own data pipeline, its own performance metrics, its own deployment readiness framework. Foundry would allow Uber to standardize how it evaluates and coordinates all of them through a single operational layer — a critical advantage as the AV transition accelerates and the cost of fragmentation compounds.

Fraud and safety operations also represent a big area. Uber processes billions of trips annually across dozens of regulatory environments. Foundry’s real-time functions, combined with AI agents capable of autonomous triage and response, could reduce fraud losses while improving incident response times at a scale that manual systems cannot keep pace with.

You don’t manage that with Tableau and a Snowflake database.

What’s also interesting is the fact that some of these companies happen to have Foundry connections, with Aurora, Waymo and NVIDIA having Foundry links of their own, meaning they are also using the platform. It’s quite exciting seeing these situations develop, especially when talking about the most important companies in their industries. Of course, Palantir finds itself in the midst of these quiet revolutions.

Uber’s core operational challenge is one that Foundry is architecturally designed to address. The company sits at the intersection of an enormous and fragmented data ecosystem: real-time pricing signals across thousands of cities, driver behavior analytics, demand forecasting models, fraud detection pipelines, AV telemetry from partners running entirely separate technology stacks, multi-jurisdictional compliance requirements, and a delivery business that runs parallel to the core ride-hailing operation. Each of these systems currently speaks a different language. How can you integrate all of this data? How can you make Waymo and NVIDIA talk to each other in a linear way? How can you generate alpha without losing any effort?

That is precisely what the Ontology does: a semantic model that maps data objects — a trip, a driver, a surge zone, an incident — into a shared structure that AI agents can act on in real time. It is not a database. It is not a dashboard. It’s a Digital Twin. It is a living representation of how an organization’s world is structured but at a speed and scale no human can match. It has been proven and it’s effective. Uber adopting Foundry was only a matter of time.

Uber is becoming a platform for autonomous mobility. Waymo rides already go through Uber’s app. More partnerships are coming. The company has positioned itself as the distribution layer for the autonomous future and even the present — the place where supply (robotaxis, human drivers and more) meets demand (people who need to get somewhere, which will always exist).

You model a vehicle object. It doesn’t matter if the vehicle is a Waymo or a Toyota Camry with a 4.8-star driver. Both are vehicles. Both connect to Zones, Trips, Demand Events. The intelligence layer sits on top and treats the fleet as one thing, optimizable as one thing.

That is the future Uber needs to build, and Palantir is how you build it. As Chad Wahlquist once told me, Palantir helps you get sh*t done.

Uber is a node in the global economy, and they are lucky to be in the hands of Palantir for their business.