Disclaimer: when a storage unit/subsystem fails, it is way better to have valid backups rather than try the recovery lottery.
That said, while modern SSDs have lower failure rate than mechanical HDDs, when they fail, they fail catastrophically. Rather than individual NAND chips, quite often is the FLT (Flash Translation Layer) or the controller which fails. In this case, recovery is extremely hard, as basically all SSDs scramble data as they are written on the NAND chips. This means that even having expensive laboratory equipment to de-solder the BGA chips, a direct reading from them is going to be relatively useless. Rather, it is a first step only to a much harder/longer recovery process.
On the other side, HDD are relatively "simpler" beast. Even the most damaged HDD (short of total disk disruption) can be read by carefully transplanting the platter onto an appropriate device to do a bit-per-bit read of the magnetic data. Note, however, that extracting consistent data is another matter - for example, CoW filesystems (as ZFS and BTRFS) themselves scramble data at an higher (ie: logical) level.