Waymo's In-House Autonomous Driving Chip Enters Mass Production, Computing Power Exceeds 1,000 TOPS Rivaling NVIDIA
Nashnova编辑部
Waymo says its in-house ASIC chip is now in mass production aboard its latest robotaxis, delivering over 1,000 TOPS — on par with Nvidia's top autonomous-driving silicon — as the company moves to control its own core compute hardware and cut costs.
How powerful is this chip?
Waymo's custom chip delivers over 1,000 TOPS (trillion operations per second), putting it in the same tier as Nvidia's latest autonomous-driving systems.
It is built on TSMC's 5 nm process — not TSMC's bleeding edge, but a mature node still widely used by Apple, Qualcomm, and other major chipmakers.
This means → Waymo chose reliability and supply certainty over the newest process, prioritizing volume production over spec bragging rights.
Why build its own chip?
Waymo's earlier fleet ran on Jaguar vehicles retrofitted with Nvidia and AMD chips. This means → the cost and delivery schedule of every car's core compute hardware depended on outside suppliers.
The in-house chip works hand-in-glove with Waymo's custom-designed sensors, processing sensor data and running AI models faster — boosting reaction time and navigation in complex urban environments.
In plain terms = building your own chip lets the "eyes" (sensors) and the "brain" (chip) work as a single system instead of a bolted-together kit.
Which vehicle gets the chip first?
The chip debuts in a new robotaxi built with Zeekr, a brand under China's Geely group.
Compared to the retrofitted Jaguar, the new vehicle is designed to be roomier and easier for passengers to use.
This reflects a broader shift: Waymo is moving from retrofitting existing cars to designing the entire ride experience around autonomy from the ground up.
How does this fit Alphabet's bigger strategy?
Waymo's parent company is Alphabet (Google's parent). Google itself is pushing custom silicon — its TPU chips — to control the cost of AI compute infrastructure.
Waymo leads the U.S. robotaxi market, operating at greater scale in San Francisco, Phoenix, and Atlanta than Tesla or Amazon's Zoox.
This means → Waymo's chip is not a standalone move. It is Alphabet's company-wide playbook — cut AI costs with in-house hardware — applied to the self-driving track.
What to watch next?
The core test: can the in-house chip keep delivering cost savings as the fleet scales from thousands of vehicles to tens of thousands?
In plain terms = making the chip is step one. The real exam comes when the fleet grows by an order of magnitude and the unit economics still have to hold.
Waymo's rivals — Tesla and Zoox — are each pursuing their own chip-to-vehicle vertical integration. The hardware-independence race in autonomous driving is just getting started.
Content is for reference only, not financial advice.