One of Samsung Foundry’s Principal Engineers has publicly announced Elon Musk’s Tesla AI5 chip is on track for mass production.
The one who first revealed the news on LinkedIn was James Kim, confirming the chip has reached tape out and will soon be manufactured at Samsung’s Taylor fab before making its way into Tesla’s newest products, adding that it had been a rewarding few months collaborating with Tesla’s engineering teams in Palo Alto and Austin.
Elon Musk first showed off an early AI5 sample in mid-April, revealing at the time that the chip would be produced simultaneously at both TSMC and Samsung Foundry. The TSMC version reportedly reached tape-out several months ahead of its Samsung counterpart.
Looking at the AI5 module’s specs, this is about half a reticle in size, and is surrounded by 12 memory packages estimated to carry a fairly wide external memory interface. Reports assumed it could be either GDD6 or GDDR7 memory, in which both are acceptable, and it also means that the AI5 could run on a 384-bit memory bus, translating to bandwidth anywhere between 768 GB/s and 1.536 TB/s.
And of course, the world’s first trillionaire claimed it could deliver up to a 40x performance improvement over its predecessor in certain workloads. He’s also said he expects AI5 to become one of the most-produced chips ever, which is the reasoning behind splitting production across two foundries.
The chip is expected to show up across Tesla’s cars, robots, and data centers.
