Elon Musk
@elonmusk
Best was working with such a great team of AI hardware & software engineers! It was more fun than going to parties on Saturdays by far.
Least awesome was that we had to make several design concessions to move fast, but were able to finish tapeout 45 days ahead of schedule.
AI6 with LPDDR6 memory addresses those design concessions and has many new great ideas. It will deliver a true doubling of performance over AI5 in the same half reticle size using the Samsung 2nm fab in Texas. AI6.5 will further improve performance using TSMC 2nm in Arizona.
Note, both chips have ~half of the TRIP AI computation accelerators dedicated to SRAM, so effective memory bandwidth is an order of magnitude greater than DRAM bandwidth for any calculations in SRAM cache.
20:04 · 15. April 2026 · 532.071 Aufrufe
325
560
5219