Must be slow news week for Digital Foundry because such a hypothetical makes very little sense.
A PS5 in 2018 would be pointless because it would still be stuck on 16nm process, meaning Sony would be limited to hell and back with regards to what they could package together for $399 / £350.
7nm process is what is needed for a proper next gen jump and 2019 (at the earliest) is the targeted date for that process. Although it is very likely it will be 2020 when we see PS5. With 7nm, Sony would be able to do an eight core Ryzen (high clocked) paired with 24GB - 32GB of HBM3 or GDDR6 (@ 1TB/s) and upwards of a 16TF GPU (perhaps more if that new chiplets design comes about by that time). Still being stuck on 16nm we're talking about an 8TF system with maybe 16GB RAM and a six core or less Ryzen (low clocked). It's just not worth it.
Especially when Cerny himself has stated that he views next gen jumps in consoles needs to be significant. That 16nm setup I just detailed would not be that. It's would be a Scorpio+ (same way that Scorpio is a PS4 Pro+).
Now, there is 12nm and 10nm process nodes, but all the big component players are choosing to skip those processes for the far more substantial 7nm (same way the big players skipped 20nm for 16nm).
They also misunderstand Cerny's statement with regards to an 8TF console is needed for 4K. Cerny gave that number to show what is needed to run PS4 games in 4K native across the board, not that he was hinting towards such a system being made. Devs would not be able to push for next gen rendering techniques AND 4K with just 8TF. It would literally just be PS4 spec games in 4K. That ain't gonna sell PS5 hardware.