Elon Musk Admits He Still Has “A Tremendous Amount To Learn” About Chip Manufacturing
Elon Musk has sparked fresh discussion around semiconductor manufacturing after acknowledging that he does not yet know how to make chips, despite previously expressing confidence that China could overcome major chipmaking restrictions within the next few years.
Speaking during a conversation with Tesla Owners Silicon Valley, Musk said he still has “a tremendous amount to learn about chip making” and clarified that he does not personally know how to manufacture semiconductors. Even so, he remained optimistic that he and the team behind Tesla’s reported Terafab ambitions can eventually figure out a path forward.
The comments come shortly after Musk told Chinese media that he believed China could solve its “compute constraints” related to lithography and chip production in roughly two to three years. That remark drew attention because advanced chip manufacturing remains one of the most complex and tightly controlled areas of the global technology supply chain.
Musk’s interest in chip fabrication has grown as artificial intelligence, autonomous driving, robotics, and high-performance computing become increasingly important to Tesla and his other companies. Modern AI systems require enormous computing power, and that demand depends heavily on advanced semiconductors. For a company pushing deeper into AI, self-driving systems, humanoid robots, and large-scale data processing, securing future chip supply could become strategically important.
The reported Terafab project has also attracted attention because of speculation about possible involvement from major semiconductor manufacturers. Musk’s latest comments suggest that outside chip companies may not necessarily lead the effort. Instead, he indicated that a company such as TSMC could potentially act more like a lessor, while Tesla or Musk-led teams would focus on developing and refining their own manufacturing approach over time.
One part of the semiconductor industry that appears to have caught Musk’s attention is the level of standardization in chipmaking equipment. Unlike Tesla’s vehicle factories or SpaceX’s rocket production systems, where custom machinery and unique manufacturing processes are common, advanced chip fabs rely on highly specialized tools that are often used across multiple industry leaders.
Musk noted that fabrication facilities operated by major players such as Samsung, Intel, and TSMC rely on similar types of equipment. His proposed approach would begin with conventional semiconductor manufacturing tools, then gradually modify them to improve wafer throughput and address bottlenecks.
According to Musk, the likely strategy would be to start with existing chipmaking equipment, identify where production slows down, and then design custom systems to increase efficiency. He suggested that the long-term goal would be to bring multiple stages of the process closer together, including mask creation, wafer fabrication, chip packaging, and repeated improvement cycles.
That kind of integrated chip production strategy would be extremely ambitious. Semiconductor fabrication is one of the most difficult manufacturing disciplines in the world, requiring cleanrooms, advanced lithography, precision chemistry, specialized materials, complex packaging, and years of process engineering. Even established chipmakers spend billions of dollars and many years developing reliable high-volume production.
Musk admitted that such a plan could take “quite a few years.” Still, his comments fit a familiar pattern: he often enters industries dominated by deep technical barriers, then attempts to reduce cost and improve output through vertical integration, automation, and rapid iteration.
The conversation also moved into artificial intelligence, where Musk shared some striking thoughts about AI-generated media and the importance of accurate training data.
Musk said he had recently seen AI systems create highly compelling music videos, suggesting that different AI tools working together could produce content with a strong emotional effect. He described this as a kind of “taste,” meaning the ability to understand what resonates with people and affects them on a deeper emotional level.
He also repeated his unusual description of experiencing “AI psychosis” or “SI psychosis” several times this year, using the phrase to describe the intense mental impact of engaging deeply with fast-advancing AI tools and AI-generated content. Musk has previously suggested that some advances in artificial intelligence may only be obvious to people who use these systems extensively.
At the same time, he warned that AI systems must be trained on accurate information and coherent principles. In his view, feeding an AI model false assumptions or forcing it to accept two incompatible ideas as true could cause serious problems in its reasoning.
“If you make it believe things that are not true,” Musk said, especially if the system is given conflicting foundational principles, it could lead the AI toward impossible conclusions. He argued that if an AI cannot reach logical consistency because its core training or instructions are contradictory, the system could become unstable in its reasoning.
Musk’s remarks highlight two major themes shaping the future of technology: the race to secure advanced semiconductor production and the challenge of building reliable artificial intelligence systems. Both areas are deeply connected. AI progress depends on powerful chips, while future chip design and manufacturing may increasingly rely on AI-driven tools.
For Tesla, the stakes are especially high. The company’s future plans depend heavily on computing, from autonomous driving and robotaxis to Optimus robots and large-scale AI infrastructure. If Musk pushes further into chip manufacturing, it could represent another major attempt to control a critical part of the technology stack.
For now, however, Musk is openly acknowledging the difficulty of the semiconductor industry. He may be confident about the long-term possibilities, but his latest comments make clear that building advanced chips is not something that can be mastered overnight.
