en.Wedoany.com Reported - Tesla's next-generation AI chip, AI6, has entered the engineering review stage. Elon Musk stated on social platform X that, considering yield rates, the usable computing power of a single AI6 wafer is expected to set a new record.

Tesla has planned multiple application scenarios for the AI6 chip, including commercial autonomous taxis, FSD software for consumer vehicles, the Optimus humanoid robot, and space data center operations. The AI6 is still in the engineering design phase, while its predecessor, the AI5 chip, has completed tape-out and is scheduled for mass production in the second half of 2027. Musk has set a nine-month development cycle for subsequent chips; based on this, the AI6 is expected to enter production in the second half of 2028.

In terms of performance, the upcoming AI5 chip will deliver five times the computing power of the two AI4 chips currently used across Tesla's entire vehicle lineup. Musk previously stated that the AI6's performance will double that of the AI5, a leap involving comprehensive innovations in processor computation and memory management architecture. Starting with the AI5, the new hardware platform will feature larger memory, as the latest FSD system has already pushed the AI4 chip's memory to its limits.

To avoid computational bottlenecks, the AI6 and its mid-cycle iteration, the AI6.5, will equip nearly half of their TRIP artificial intelligence computing accelerators with static random-access memory (SRAM). This high-speed on-board memory allows the processor to perform complex computations in the cache, without waiting for the system's main memory to respond. For main memory configuration, the AI6 will adopt the faster sixth-generation low-power double data rate memory (LPDDR6), further improving performance over the LPDDR5 and LPDDR5X memory used in the AI5 architecture.

Preparations for chip mass production are gradually advancing. Tesla is deeply collaborating with Samsung, which will manufacture the AI6 chip at its new semiconductor factory in Texas, with the chip foundry cooperation agreement valued at $16.5 billion. Additionally, Tesla is partnering with Intel and SpaceX to advance the TERAFAB artificial intelligence chip project, aiming for full independent integration of the semiconductor production chain.

Car owners will not see this new chip in production vehicles in the short term. Musk stated that the next-generation AI chip will first be applied to the Optimus robot and the supercomputer cluster used for training Tesla's neural networks, before gradually being deployed to consumer vehicles. Musk believes that the current AI4 chip's performance is already sufficient to achieve driving safety levels surpassing humans, giving the hardware team ample time to refine and optimize the AI6 before officially integrating it into vehicles.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com