Elon Musk Shares First Images of Tesla AI5 After Tape Out as Tesla Pushes Toward AI6 and Dojo 3
Elon Musk has revealed the first public images of Tesla’s next major AI processor after confirming on X that the company has successfully taped out its new chip. One important correction comes first: this is AI5, not A15. Musk’s post specifically congratulated the Tesla AI team on taping out AI5 and added that AI6, Dojo 3, and other chips are already in development. Coverage from multiple outlets tracking the post also matches that wording.
In the image Musk shared on X, the package appears to feature a large central compute die surrounded by 12 memory packages. Several reports analyzing the photo identified those memory chips as SK hynix parts, though the exact memory SKU is not publicly confirmed from Tesla itself. Tom’s Hardware also noted that the package appears to use standard external memory rather than stacked HBM, which is a meaningful design detail for cost, manufacturing scale, and packaging strategy.
And thank you to @TaiwanSemi_TSC and @Samsung for your support in bringing this chip to production! It will be one of most produced AI chips ever.
— Elon Musk (@elonmusk) April 15, 2026
Tesla is framing AI5 as the successor to Hardware 4, and Musk has previously described it as a massive leap over AI4. Reported comments from Tesla’s earnings discussions indicate he claimed AI5 could be up to 40 times better than AI4 in certain scenarios, though that is Musk’s own framing rather than an independently verified benchmark suite. The same reporting also suggests AI5 is intended not only for Tesla vehicles, but for Optimus and broader Tesla AI workloads as well.
That said, several of the more detailed specifications circulating online should still be treated with caution. Claims around roughly 2500 TOPS, 144GB of memory per chip, and direct performance comparisons to NVIDIA Hopper or Blackwell are not confirmed in the sources I could verify today from Tesla or Musk’s new post. Some of those figures have been repeated by secondary coverage, but they do not appear in the confirmed tape out announcement itself. The safer conclusion is that Tesla has reached a major chip design milestone and has shown the package for the first time, but many of the fine grain specifications remain unconfirmed publicly.
The road ahead is also becoming clearer. Musk’s own message says AI6 and Dojo 3 are already in the works, which means Tesla’s silicon roadmap is moving beyond AI5 even before this new chip reaches volume deployment. Reports tied to Tesla’s January 2026 commentary also indicate Musk had restarted work on Dojo 3 after AI5 reached a more stable stage, reinforcing that Tesla sees its in house compute platform as a long term pillar for autonomy, robotics, and data center ambitions.
Solving AI5 was existential to Tesla, which is why I had to focus both the teams on that chip and I’ve personally spent every Saturday for several months working on it.
— Elon Musk (@elonmusk) January 19, 2026
This will be a very capable chip. Roughly Hopper class as single SoC and Blackwell as dual, but it costs…
On the manufacturing side, Musk thanked both Taiwan Semiconductor and Samsung in connection with bringing AI5 to production, which lines up with current reporting that Tesla is working across multiple major semiconductor partners. At the same time, Reuters reported today that Musk’s team has begun reaching out to suppliers for the Terafab project, a proposed chip complex tied to Tesla and SpaceX, with the goal of eventually starting silicon manufacturing in 2029. That does not mean Terafab is ready now, but it does show that Tesla is thinking far beyond a single chip generation and wants deeper control over future production capacity.
From a strategy standpoint, AI5 matters because it is not just another automotive chip refresh. Tesla is increasingly positioning its custom silicon as a foundation for a wider physical AI stack that spans self driving, humanoid robots, and larger compute infrastructure. If AI5 lands close to Tesla’s internal expectations, it could become one of the company’s most important hardware platforms of the next few years. But the biggest immediate takeaway is more grounded: Tesla has completed tape out, shown the chip publicly, and confirmed that AI6 and Dojo 3 are already moving forward. Everything beyond that now depends on fabrication, validation, yield, software readiness, and how quickly Tesla can turn this milestone into shipping products.
Do you think Tesla’s custom AI chips can become a real alternative to the broader NVIDIA centered AI hardware ecosystem, or will AI5 still be too specialized to change the bigger market?
