Autonomous Vehicles Get Buff as WeRide Injects Them with Dual-Thor Power
WeRide just launched its HPC 3.0 platform—a dual-NVIDIA Thor chip-powered brain for Robotaxis that cuts costs by 84%, runs 2,000 TOPS of compute, and meets more certifications than a pilot’s license.

Remember when autonomous driving meant you might one day nap in the backseat while your car responsibly merged onto the freeway? Well, forget that. Now it means your Robotaxi comes with not one, but two NVIDIA Thor chips, enough AI horsepower to run a mid-sized moon base, and the manufacturing grace of Lenovo, your favorite laptop battery's third-degree cousin.
Today, WeRide — the world’s first publicly listed Robotaxi company, because of course that’s a thing now — dropped what they call a game-changer: the HPC 3.0 platform. What does that mean for you? Mostly, that your next rideshare might be powered by more TOPS than your entire office IT department, all while costing less than an espresso machine at WeWork.
Let’s break it down.
💾 The Chip That Drives You (Literally)
WeRide’s HPC 3.0 is powered by NVIDIA DRIVE AGX Thor, the chip equivalent of Thor’s hammer, if Mjölnir had safety certifications and could operate at 85°C in a sandstorm. This system pumps out 2,000 TOPS of AI compute — or as your VC cousin says, “enough to run fifteen LLM startups and still mine Dogecoin in the background.”
And yes, it runs DriveOS, which we’re assuming is like Windows for robots, except it probably crashes less often.
🛠️ Built Like a Tank, Except It’s a Taxi
According to WeRide, this bad boy meets every acronym known to the auto industry: AEC-Q100, ISO 26262, IATF 16949. We don’t know what those mean either, but it sounds very certified. They even brag about a failure rate under 50 FIT — which is either a measure of safety or how often a human passenger screams “JESUS TAKE THE WHEEL.”
It can also operate from -40°C to 85°C, which is great if you live somewhere between Antarctica and the surface of Mars.
🧮 Do the Math, Then Burn It for Efficiency
The press release proudly declares that HPC 3.0 reduces autonomous driving suite costs by 50%, production costs by 75%, and TCO (total cost of ownership) by 84%. Which, if the trend continues, means the HPC 4.0 will be free and the car will pay you to ride it.
🚍 Coming Soon: RoboEverything
The HPC 3.0 platform will power not just Robotaxis, but also WeRide’s Robobus and Robosweeper, because even your street-cleaning vehicles need AI now. Expect generative sidewalk etiquette and GPT-powered horn honks in a city near you.
🤝 Lenovo, NVIDIA, and the Holy Trinity of Autonomous Bragging Rights
WeRide teamed up with Lenovo (famous for laptops with 17 status lights) and NVIDIA (famous for making the GPUs you still can't buy). The result is a car that probably has more GPU power than your data center and more press quotes than ChatGPT had updates this month.
🏁 Final Snark
This isn’t just a car. It’s an edge-compute, thermally certified, multi-standard-validated, GPU-bonded rolling SDK. And it's coming to streets that already can’t handle e-bikes.
But hey — it’s 84% more cost-efficient. So if the robots are going to take the wheel, at least they’re doing it on sale.