It looks like Nvidia messed up, or maybe they planned it this way
https://www.igorslab.de/en/chipsize-leaked-vidia-ampere-ga102-in-sams-8-nm-lpp-always-more-likely/
Samsung's "8nm" process is really nowhere close to being as efficient as TSMC's 7nm or 7nm+.
That's why Ampere is so power hungry with several high end cards featuring over 300w TDP specifications.
Perhaps Nvidia got cocky and decided to take the cheap option so they could maximise margins.
However if Nvidia manages to still stay ahead of AMD with a vastly inferior process again, that will be very interesting indeed. On the flip side, if AMD wins the next generation - it will be a big egg on Jensen's face for taking the cheap option to go with Samsung instead of the industry leading process node from TSMC.