Before assuming the drive is dead, spend five minutes ruling out the machine. M.2 slots work loose, SATA cables fail, and firmware hides drives — all of which look identical to a failed SSD from where you are sitting. When those are exhausted, here is what the symptom actually means.
$ cdr diagnose /dev/nvme0 → Device: Samsung 970 EVO · 1 TB · NVMe → Status: NOT DETECTED — controller failure → Client: confidential · Ely, Cambridgeshire $ cdr engineer-working → NAND read: chip-off · raw flash dumped → Reconstruction: wear-levelling + ECC resolved → Imaging: 0.91 TB / 1.00 TB (91%) $ cdr verify → ✓ documents — 71,440 files → ✓ photos — 44,900 files → ✓ data reconstructed — recovered
The controller has stopped responding. Most common by a distance, and the NAND behind it is usually intact. Reached through vendor mode or by reading the memory directly.
The drive has lost or corrupted the tables describing its own geometry. A firmware-class fault rather than physical damage, and often among the more recoverable presentations.
The hardware is working and the file system is not. The cheapest class, and the one most often made worse by accepting Windows' offer to initialise or format.
Where the controller no longer responds, there are two ways to the data. The first is vendor mode — a manufacturer-specific engineering interface that allows the controller's own firmware and translation tables to be examined and repaired. This is the preferable route because the drive reassembles its own data.
Where the controller is genuinely dead, the NAND is read directly. The memory chips are accessed and the raw flash extracted, then the data reassembled by reversing what the controller was doing to it — wear levelling, interleaving, XOR scrambling and error correction, all of which vary by manufacturer and firmware version.
That reassembly is the difficult part and it is why SSD work is not a matter of plugging the chips into something. Without correctly reversing the controller's arrangement, raw NAND is noise.
SSD and NVMe recovery is from £300 +VAT. Vendor-mode work and direct NAND reads are priced at the same class — the figure is set at diagnostic by fault type, not by how involved the afternoon turns out to be.
No fix, no fee applies to nearly all solid state work, because there is rarely a committed parts cost of the sort clean-air mechanical work carries.
Because solid state gives none. SMART reports flash wear reasonably and controller health barely at all, and it is the controller that usually fails. There was no sign to miss.
In the great majority of cases, yes. The failure is in the controller or firmware rather than the memory, so the data is physically present and the problem is reaching it.
No. Initialising writes a new partition structure and destroys the one describing your files. That prompt appears whenever Windows cannot read a disk, which is exactly what a recoverable fault looks like.
Considerably. Several NVMe drives enable encryption by default. With the key or password it makes no difference; without either, there is no route in regardless of how healthy the NAND is.
On most Windows laptops, yes — the flash is read at chip level. On Apple T2 and Apple Silicon machines, no, because the key is fused into a security chip on that same board.
Three to ten working days. Vendor-mode work is quicker; direct NAND reads take longer because the controller's arrangement of the data has to be reversed before anything is readable.
Controller, firmware and chip-level recovery, all in-house, a free written diagnostic to open, and no fix no fee on most jobs. Begin your recovery today.