Image: Tesla
Elon Musk has laid out Tesla’s most ambitious AI roadmap yet — one that stretches far beyond self-driving cars and humanoid robots, and eventually into space.
In a recent post on X, Musk broke down how Tesla’s next several generations of in-house AI chips will be used across the company’s expanding ecosystem. “AI4 by itself will achieve self-driving safety levels very far above human,” Musk wrote about the hardware Tesla’s EVs are currently using. “AI5 will make the cars almost perfect and greatly enhance Optimus. AI6 will be for Optimus and data centers. AI7/Dojo3 will be space-based AI compute.”
The comments build on Musk’s growing stream of updates about Tesla’s custom silicon efforts. Earlier this week, he said the next-generation AI5 chip — also known as Hardware 5 (HW5) — is “almost done,” with work on AI6 already underway. AI5 is expected to deliver up to a 50x improvement over today’s AI4 hardware, forming the backbone of Tesla’s future Full Self-Driving system, Robotaxi fleet, and Optimus humanoid robots.
While AI4 already powers today’s FSD (Supervised), Musk has repeatedly claimed that AI5 will push Tesla vehicles to safety levels “very far above human,” a key requirement for large-scale autonomy. However, production timelines remain long-term. Musk has previously said that Tesla is targeting 2027 for volume production of AI5, with the chip not making its way into production vehicles until the middle of the year. AI5 will be manufactured in the U.S. by both TSMC and Samsung.
AI6, expected to follow in 2028, will take things further. Tesla has already signed a massive $16.5 billion deal with Samsung to produce the AI6 chip domestically, and Musk has described it as potentially “the best AI chip by far.” Unlike AI5, which is primarily vehicle-focused, AI6 is being designed with Optimus robots and large-scale data center compute in mind.
Perhaps most striking is Musk’s long-term vision for AI7, also referred to as Dojo 3. According to Musk, this generation will be “space-based AI compute,” hinting at orbital AI infrastructure that could support everything from global autonomy networks to satellite-based processing. Musk also noted this week that Tesla is targeting an aggressive nine-month design cycle for its custom silicon, suggesting a rapid-fire release schedule for future AI chips.
With Tesla simultaneously scaling Robotaxis, humanoid robots, and AI training systems, it’s becoming increasingly clear why Musk has floated the idea of building a massive chip fabrication facility of Tesla’s own. Between cars, robots, data centers, and eventually space-based compute, Tesla’s appetite for AI silicon is only just beginning.
