American Made Warehouse Drones: Why We Design and Manufacture the Full Stack in Mountain View, California

Corvus Robotics makes its own hardware and its own software. We design the full stack in America, from the sensor and compute architecture to the AI models running on each device. Every Corvus One® and Corvus Trident™ unit is manufactured and tested at our Mountain View, California headquarters.

Our products live indoors. You’ll find Corvus One working the aisles of busy warehouses, distribution centers and manufacturing plants (freezers at minus 20F included), while Corvus Trident rides on the forklifts and reach trucks driving past it. The drone flies around without GPS, WiFi or anything attached to the racking or ceiling, so no stickers, beacons or markers. Since both systems run on vision, whatever they see gets processed right there on the device. Both are ours, from the first design to the finished unit.

Why robotics companies have to build their own hardware

What actually holds a robot back? In our experience it’s the hardware, pretty much every time. Code can only be as good as the machine running it, which means a perception model is stuck with whatever image the sensor delivers and a drone that needs to correct course is waiting on its onboard processor.

Think about what a buyer wants to know about a warehouse drone. Can it get down a tight aisle, react in time, read a label that’s frosted over at minus 20F and stay up long enough before it has to land? Those answers get fixed the moment someone chooses the airframe, the sensor and the compute module, which happens long before a single line of inventory code exists. Everything written afterward has to work inside that box. So we’d argue the American robotics industry is going to rise or fall on hardware. Give it five years and people will wonder why this was ever up for debate.

Our founders learned it the hard way. They’re roboticists, and they spent the early days building prototypes, testing them, pulling them apart and rebuilding them. Pretty quickly it was clear the system wouldn’t survive a real warehouse unless we controlled the hardware as well, and that decision is why our drones are working at enterprise scale right now.

The risks of software-only robotics

To give the other camp its due, skipping hardware is a defensible call. The idea is to build a great vision system, put it on another company’s commodity drone and let the software make up the difference. Customers don’t have to budget for capital equipment and the vendor never needs a factory. And if a better airframe hits the market, switching costs them nothing in tooling or retraining.

The trouble starts when those vendors try to improve. In 2025, one of them announced a big step forward: tighter aisles, the ability to dodge moving obstacles and depth data precise enough to infer case counts. All of it came from new hardware, a logistics variant that a drone platform company co-developed for them. The new aircraft joined a fleet that was DJI, and the announcement made a point of calling it American made (which tells you what else was pushing the change). So a software company had to work with a partner to build it a new drone, on the partner’s timeline, because conditions had shifted under the one it was utilizing throughout its fleet.

Cold storage makes the problem even more obvious. A drone has to hold its position at minus 20F while frost creeps over the very labels it’s trying to scan, and that takes an airframe engineered for the job. Somebody covers that engineering bill, whether it shows up in the vendor’s costs or the customer’s quote.

Owning only the software means accepting choices made by someone else. The vendor can put in requests (partners do co-develop hardware, so it has some pull), but those requests get prioritized against another company’s budget and production calendar, and the final product has to serve that company’s other customers as well. Specs, pricing and availability can shift overnight. Even the devices gathering the data the vendor sells, on the rack or on the forklift, came out of another company’s shop.

Then the federal government weighed in. In December 2025, an executive branch interagency body made a national security determination about foreign-made unmanned aircraft and their critical components. The FCC acted the next day, adding foreign-produced drones and UAS critical components to its Covered List. Action against DJI and Autel was the obvious move, but the list went further and swept in every foreign-produced airframe, which caught a lot of the industry off guard. In practice the impact is limited, although it’s real. A covered device can’t receive a new equipment authorization, meaning new models can’t be imported, marketed or sold in the United States under that authorization. Aircraft authorized before the listing are still flying. Exemptions have been extended more than once, but none of them has ever reached a DJI or Autel product. Both companies are challenging the listing. So far neither challenge has been decided on the merits, and fights like this tend to drag on for years.

Tariffs arrived in August 2026. The White House proclamation hit imported drones and drone components, and the aircraft used in warehouses fall under it. The standard rate is 25%. A few allied countries, including Switzerland, get a lower 15% rate, tied to where their hardware, software and technology come from. Since a proclamation set those numbers, another proclamation can change them, so the exposure is a bigger issue than whatever the rate is today.

Our drones come off the line in Mountain View. A customer who orders ten more gets ten more that we built. When trade policy swings, our answer holds, because there’s no supplier to requalify.

Corvus One video transcript ▾

Meet Corvus One, your autonomous inventory management system. Fully autonomous, lights-out data capture. Powered by third-generation Corvus intelligence, trained on millions of interactions from the real world. Scan barcodes, find empties, and count cases. Works any shift, day or night. No pilot, reflectors, or stickers for navigation.

What purpose-built hardware makes possible

When one team designs the aircraft and the software together, scans get faster, the data gets cleaner and more barcodes get read than if you’d bolted software onto another company’s drone or camera. Take the flight control loop. It runs on compute we pick and closes in less than 10 milliseconds, which is the margin between spotting a rack upright and clipping it. We applied the same logic to Corvus Trident, a camera system we built to ride on material handling equipment and read pallets without the truck ever stopping. No other company’s roadmap was going to produce that for us.

Controlling the hardware lets us build for what warehouses are really like. Lighting is dim, aisles are tight and dusty, and traffic keeps moving while the drone works. Freezers raise the difficulty again, with air at minus 20F, frost on the labels and blowers churning the whole time. Batteries are their own problem in there (cold cells lose capacity and won’t charge properly), so running in a freezer around the clock meant engineering the pack and its thermal management around low temperatures. Each of these gets solved in the hardware, whether it’s the optics, the battery pack or the way the airframe holds steady in moving air.

For us, that also means nothing has to be installed in the building for the drone to navigate. That isn’t the case everywhere. Some off-the-shelf airframes depend on navigation stickers applied to the racking, and other systems need battery-powered beacons placed around the facility, which then have to be maintained and relocated every time the layout changes. In both cases the building turns into a piece of the system, with every sticker and beacon a new place for something to fail. We own the perception stack and the hardware under it, so we engineered that requirement away entirely.

Reliability, security, and data integrity

A vendor that sources its hardware can only tell customers what the supplier told it about how data gets captured and where it ends up. Ownership of that data after it lands is governed by the supplier’s terms. And data handling was a big factor in foreign-made airframes landing on the Covered List in the first place.

We account for every board inside Corvus One and every place its data travels. Customers choose whether it goes to the cloud or stays on premise.

Now picture a drone that starts drifting, or crashes, or does something no one saw coming. A software-only vendor opens a ticket with its manufacturer and waits for a fix on the manufacturer’s timeline, and that fix might not show up until the next product generation, two or three years down the road. Custom work is harder still, because now the vendor is negotiating with its supplier on the customer’s behalf. Because we own the hardware, our fix lands on the device itself.

Certifications and standards

An independent registrar certifies Corvus Robotics to ISO/IEC 27001, the international standard for managing information security. TÜV Rheinland certifies the Corvus One drone battery to UL 2054, and the battery is also tested to UN 38.3, the United Nations standard for transporting lithium batteries safely. For the European market, the system carries CE marking and RoHS compliance, both declarations of conformity that we issue as the manufacturer. Corvus Robotics also has SOC 1 and SOC 2 reports covering its information security controls, both attested by an independent third party.

All of this happens in one place. Every Corvus One and Corvus Trident is designed, manufactured and tested in Mountain View, California. That’s how a drone ends up working a freezer aisle at minus 20F in a dark, empty building, and it’s how we can localize a forklift or reach truck precisely enough to know the full route each pallet takes after it’s picked up. Whatever problem a customer hits next, we’ll be the ones solving it, on our own schedule.