Elon Musk Says Tesla AI5 Chip Will Use 96GB RAM as Company Adjusts Plans for Optimus and Self-Driving AI
Elon Musk has shared a new update on Tesla’s upcoming AI5 chip, confirming that the processor will now use 96GB of LPDDR5 memory instead of the previously mentioned 72GB. The change comes as Tesla continues preparing the hardware that will support its next generation of autonomous driving systems and the Optimus humanoid robot.
The AI5 chip is a key part of Tesla’s long-term artificial intelligence strategy. Musk has repeatedly described autonomy and robotics as central to the company’s future, and AI5 is expected to play a major role in both areas. The chip reportedly taped out in April, marking an important step toward production.
Musk previously said Tesla had reduced memory capacity for its AI5 and AI6 chips because of tight global memory supply. According to him, the company had cut AI5 memory to 72GB of LPDDR5 and AI6 memory to 144GB of LPDDR6 in order to improve the chances of securing enough components during a difficult supply environment.
However, Musk later clarified that Tesla decided to raise AI5’s memory configuration to 96GB. He explained that the company did not want to be the only major customer using the minimum LPDDR5 memory version. By increasing the capacity, Tesla appears to be balancing supply concerns with performance requirements for advanced AI workloads.
The update is especially important because memory capacity is a major factor in AI performance. Tesla’s AI systems must process large amounts of real-time data from cameras, sensors, vehicle systems, and robotics applications. More memory can help support larger models, faster decision-making, and better overall efficiency in self-driving and humanoid robot tasks.
Musk has made bold claims about AI5’s potential. He has suggested that Tesla’s chip could compete with NVIDIA’s Hopper-class AI hardware on a standalone basis and approach Blackwell-level performance when used in a dual-chip setup. He has also claimed that AI5 could offer significantly better performance per watt, which would be crucial for vehicles and robots where power efficiency matters.
According to Musk, Tesla’s advantage may also come from cost. He has previously said the AI5 chip could deliver competitive AI performance at only a fraction of the cost of comparable data center-focused hardware. If Tesla can achieve that, it would strengthen the company’s ability to scale artificial intelligence across millions of vehicles and future Optimus robots.
Alongside the hardware discussion, Musk also commented on the rapid progress of AI in programming. Responding to a user who suggested that the best way to understand how quickly AI is improving is to use it regularly for coding, Musk agreed with the point. The comment also noted that frequent AI use could make it harder to explain those advances to others, because the pace of improvement becomes more obvious only through direct experience.
The discussion touched on the phrase “AI psychosis,” a term that has gained attention as experts debate the psychological effects of prolonged chatbot use. Some mental health professionals have warned that AI systems may reinforce certain beliefs in vulnerable users, especially when interactions lack grounding in real-world feedback. Musk’s brief response did not expand on the issue, but it reflects growing concern about how deeply AI tools are becoming integrated into work, communication, and problem-solving.
Tesla’s AI5 update shows how complex the race for AI hardware has become. The company is not only competing on raw performance but also managing memory supply, power efficiency, cost, and scalability. With Optimus development and autonomous driving still central to Tesla’s roadmap, the AI5 chip could become one of the company’s most important technologies in the years ahead.
