Elon Musk Upgrades Tesla AI5 Chip Memory to 96GB
Strategic Hardware Adjustments
Tesla Chief Executive Officer Elon Musk clarified that the upcoming AI5 chip will feature 96GB of memory instead of the previously planned 72GB of DDR5. This strategic adjustment prevents Tesla from becoming the sole manufacturer utilizing the minimum 72GB configuration standard. This specific silicon will notably power the highly anticipated Optimus humanoid robot.
Musk previously indicated that Tesla halved the memory specifications for both AI5 and AI6 chips. This difficult decision aimed to secure reliable component supplies amidst a historic global memory shortage. During these profound reflections, he also acknowledged a fascinating psychological phenomenon. He noted that relying excessively upon artificial intelligence for programming might induce a form of “AI psychosis.” Once this cognitive state occurs, developers find it incredibly difficult to articulate actual AI advancements to outside observers.
The Core of Tesla Robotics and Autonomy
The revolutionary AI5 chip represents the fundamental core of Tesla’s autonomous driving and robotics endeavors. Musk firmly believes these two critical business sectors will ultimately dictate the future trajectory of the entire company. The executive proudly confirmed that the AI5 chip completed tape-out in April. Consequently, he harbors immensely high expectations for this robust hardware. He asserts that a single AI5 chip rivals the performance of an NVIDIA Hopper AI graphics card. Furthermore, a dual-chip configuration can effectively compete against the more advanced NVIDIA Blackwell architecture.
Musk also claims that the AI5 chip delivers three times the performance per watt compared to competing NVIDIA processors. In late August, Musk revealed his intimate familiarity with the intricate chip design. Moreover, he stated that the overall manufacturing cost is merely ten percent of comparable NVIDIA products, assuming equivalent performance metrics.
Securing Hardware Amidst Supply Shortages
Yesterday, Musk divulged further compelling details regarding the AI5 chip architecture. He explained the initial rationale behind halving the LPDDR5 memory capacity. Writing on the X platform, he stated: “We halved the memory for the Tesla AI5 chip (originally 72GB of LP5), and reduced AI6 memory by a third (currently 144GB of LP6).” This significant capacity reduction stemmed directly from persistent supply constraints within the global memory market. Lowering the hardware specifications served as the only viable method to guarantee a stable supply chain.
However, Musk shortly thereafter provided an important addendum. The AI5 will no longer utilize the 72GB configuration, upgrading instead to 96GB. According to his explanation, Tesla decided to slightly increase the AI5 memory to 96GB. Otherwise, Tesla would remain the only enterprise utilizing the absolute lowest tier of this specific LP5 memory variant.
The Phenomenon of AI Psychosis
Simultaneously, Musk shared his visionary perspectives regarding overarching artificial intelligence development. Last month, Musk frankly admitted that AI is currently insufficient for developing highly demanding, performance-critical software. However, he simultaneously predicted that AI will acquire this profound capability very soon.
Previously, an X platform user commented on the rapid iteration speed of artificial intelligence. The user noted that frequently utilizing AI for programming is the only way to truly comprehend its rapid evolution. Yet, this intense, frequent usage easily induces the aforementioned “AI psychosis.” This psychological barrier prevents users from effectively explaining AI breakthroughs to the outside world. Responding to this insightful comment, Musk expressed his complete agreement, simply replying, “Indeed.”
Interestingly, the concept of “AI psychosis” began garnering significant mainstream attention last year. At that time, a psychiatrist from the University of California issued a stark public warning. They cautioned that advanced chatbots could potentially trigger psychotic episodes in users already predisposed to such mental illnesses. The doctor explained that when reality conflicts with a user’s internal expectations, artificial intelligence cannot help users correct their cognitive frameworks. Consequently, this digital echo chamber renders susceptible individuals far more vulnerable to developing severe psychological complications.
