IBM demonstrated decades ago that they could position individual atoms. The question isn't whether SMIC can produce at 7nm, 5nm or even 2 nm. The real question is whether they can do it in volume, at reasonable prices.
DUV (the old technology) uses 193 nm UV light. To achieve resolutions smaller than that, the most common trick is to just repeatedly use the same mask, shifted over a few nanometers. Initially this was done twice, then three times, and by 14 nm companies were doing 4 shifts. And these 4 shifts needed to be done for every layer on the chip, so it really added up.
Now there's no technical limit to 4 steps. Just like razorblades, you can go to 5 steps, it's just getting a bit silly. SMIC might have even used 7 or 8 steps to achieve 7nm; we can't tell.
Note that in addition to the sheer time it costs, you also have defect risks with this multi-patterning. We don't know how many of those "7nm DUV" chips actually work. This obviously impacts the price; if you have to discard 90% of the dies then the remaining 10% become ten times as expensive.