Waymo Opens the Trunk: Inside the Custom Compute Running Its Robotaxis

Most of us think about what goes in the trunk in terms of cargo liners, jump starters, and folding storage bins. Waymo thinks about it differently — because for its self-driving fleet, the trunk is where the brain lives. In a new post titled "A look under our trunk: what's in our compute," the company gave its first public look at the hardware that turns raw sensor data into real-time driving decisions.

It's a genuinely interesting read for anyone who cares about EV tech, and it doubles as a reminder of just how much computing power is now riding around in the back of an electric vehicle.

Data-Center Performance, Squeezed Into a Car

Waymo describes its compute as "a state-of-the-art system that would be considered impressive for a data center, with the added complexity of an in-vehicle operating domain and real-time requirements." Unlike driver-assist systems that assume a human is ready to take over, Waymo's stack handles the entire driving task on its own — which the company says demands significantly higher performance.

That design is grounded in more than 200 million miles of fully autonomous experience, and it's built around three requirements Waymo calls non-negotiable:

• Responsive. Everything runs onboard, with decisions processed in milliseconds. Waymo says it has engineered the stack for ultra-low latency — minimizing the delay from "first pixel to action" — and has scaled raw compute power 20x in just eight years.
• Ruggedized. The hardware lives with constant vibration, shock, and extreme temperatures. It integrates directly with the vehicle's liquid cooling system to hold peak performance through "freezing Midwest winters or the blistering heat of Phoenix."
• Redundant. With no human backup, the compute is designed "like two independent engines" running full parallel workloads. If one faults, the other seamlessly takes over.

"Compute is the brain of the Waymo Driver, translating raw sensor data into real-time driving commands." — Waymo

A Purpose-Built 5nm ASIC

The headline hardware news is Waymo's custom 5nm ASIC — described as a specialized ML powerhouse engineered to process, fuse, and run neural networks on raw sensor data in real time. Its accelerators pull critical information straight out of lidar, radar, and camera streams, including temporal denoising for better low-light perception, before feeding a purpose-built inference engine for sensor fusion.

The number that stands out: these ASICs alone deliver over 1,000 TOPS of ML performance dedicated to front-end processing. Waymo notes it's just one of several custom components in development, and that the chip works inside a broader "ML-primary" architecture that pairs custom silicon with off-the-shelf CPUs, GPUs, and accelerators to handle orchestration, data movement, and logging.

Why the Trunk Detail Actually Matters

Here's the part that should resonate with anyone who has ever played Tetris with grocery bags in a Model Y: Waymo specifically calls out that its highly integrated system delivers all that processing power "without compromising the rider experience, maximizing battery efficiency while preserving ample trunk space and running silently."

That's the same tension every EV owner negotiates. Cargo space, range, and cabin quiet are finite resources, and everything you add competes for them. Waymo's engineering answer — tighter integration, shared liquid cooling, purpose-built silicon instead of stacked off-the-shelf boxes — is the industrial-scale version of the choices we all make when outfitting a car.

What This Signals for EV Owners

Waymo's post isn't a consumer product announcement, and it doesn't change anything about the car in your driveway today. But it does mark a clear direction of travel for the industry:

• Custom silicon is the new normal. Waymo moved from off-the-shelf components to purpose-built systems because generic hardware couldn't meet the latency and efficiency targets. Expect that logic to keep spreading across the EV space.
• Thermal management is a first-class problem. Tying compute into the vehicle's liquid cooling loop is a reminder that heat, not raw chip specs, is often the real ceiling on sustained performance.
• Efficiency is a compute story now. Waymo explicitly frames its integration work as protecting battery efficiency. As vehicles get smarter, the power budget for computing becomes part of the range conversation.

For now, the practical takeaway for the rest of us is simpler: your trunk is prime real estate, and how well you use it matters. Whether that means a proper cargo organizer, a well-secured emergency kit, or a charging cable setup that doesn't turn into a tangle every time you open the hatch, the principle Waymo is working from applies at any scale — integrate thoughtfully, and you get more out of the same space.