NVIDIA SHOULD BE WORRIED.
Elon Musk just posted yesterday that Tesla's AI6 chip could set a world record for the most usable intelligence from a single wafer when factoring in yield.
Every AI chip wastes a large portion of the silicon wafer it is made on due to manufacturing defects. Nvidia's chips are so large that even small defects destroy the entire chip. Tesla's AI6 is designed so efficiently that almost every chip cut from the wafer actually works, meaning Tesla gets far more usable computing power from the same amount of silicon.
More usable chips per wafer means lower cost and higher performance at the same time.
Tesla was one of Nvidia's biggest customers. That is changing fast and could be replaced completely.
In April 2026, Tesla completed its AI5 chip, which already matches Nvidia's H100 performance for Tesla's specific workloads. AI6 tapeout is scheduled for December 2026. AI7 is already being planned. Tesla is releasing a new chip generation every 9 to 12 months.
Tesla signed a $16.5 billion deal with Samsung to build AI6 chips in Texas and is also working with TSMC. It now has two of the world's most advanced chip factories competing to manufacture its silicon.
This is exactly what Apple did. Apple spent years buying Nvidia chips, built its own M1, and never bought Nvidia again. Tesla is running the same playbook, building its own chips for its cars, for Optimus robots, and for its data centers.
And Tesla is not alone. Google, Amazon, Microsoft, Meta, and OpenAI are all building their own chips to replace Nvidia.
When every major customer is designing their own silicon, Nvidia's biggest risk is not competition, it is abandonment.