[pdpi]

The PCB has plenty of space for an M.2 slot (and, in fact, the PCB has pads for the connector and mounting holes, and is appropriately silk screened). The case design is just not designed for user serviceability, so having the slot would not be particularly useful. Jeff Geerling talks about this to some length in his video.

Arguably, a non-user serviceable RPi is a bit of an odd choice, but the 400/500 are very much self-contained all-in-one products, which I guess makes it kind of reasonable.

[sho_hn]

Sure, but I'd expect a little lid to access the M.2 slot in the case design. "It's not that kind of product", but maybe it should be: Higher capacity microSD cards are more expensive than NVMe alternatives, which are generally faster.

[Ekaros]

Not to forget write cycles... Which can be real issue if you actually use system as desktop. In general NVME drives would likely be more reliable in long run.

I really would not enjoy storage being corrupted or dead on this kind of device.

[cdt5050]

A huge reason SD cards have been unreliable in the past is because of a lack of wear leveling functionality, and architecture assuming sequential writes (e.g. large buffers that require larger blocks of data to be erased and rewritten even when writes are small as is typical in random writes, resulting in excessive write cycles).

Modern SD cards have much more intelligent controllers that perform wear leveling and work with the OS to reduce excessive write cycles.

Not that they're as reliable as modern SSDs, but I'd be willing to bet a modern SD card from a reputable brand is as reliable as SSDs were a while back when they were generally regarded as reliable enough.

Almost every time I've teased the info out of people with failed SD cards on Pi's, either it's either physically damaged, or it's an off-brand, or it's an ancient card.

[whitehexagon]

I use a Pi400 daily but boot it from a small USB stick. So no SD card needed, and it seems faster.