PAK *ja-c̣ʷa (with a frequent pronominal prefix); PAK *ʁʷǝč̣ǝ́. The PAK and Ub. form reflect a PWC compound *ʁ́Iʷǝ-ƛ̣ʷV, where the first part is *ʁ́Iʷǝ 'copper, metal (in general)' (q.v.).
The need to see a colour name in the second part of the AK and Ub. compound is obvious: both in AK and Ub. the name of iron forms part of an opposition - in AK *ʁʷa-pʎá 'metal' + 'red' = 'copper, red copper' : *ʁʷǝ-č̣ǝ́ 'metal' + '?' = 'iron'. In Ub. wǝ-sʷá 'metal' + 'white' = 'copper, bronze' : wǝ-c̣ʷá 'metal' + '?' = 'iron'. If we assume that the Ub. and AK words for 'iron' are genetically related (certainly, the theory of Ub. wǝc̣ʷá being an Adygh loanword, see Shagirov 1, 140, can not be considered seriously), the correspondence PAK *č̣ : Ub. c̣ʷ is unambiguously pointing to the PWC lateral labialised *ƛ̣ʷ, for which the expected PAT correspondence is *c̣ʷ - and we find it in PAT *ja-c̣ʷa 'blue'.
The concept of iron being a "blue metal" is by no means strange: it goes back to one of the earliest varieties of tempered iron (steel), which was in Greek called kúanos (cyan) and was dark-blue in colour (on the etymology of this word and the related Hitt. kuwanna- see under *kwɨ̄wV). We must also mention the widespread Caucasian myth about the birth of the cultural hero from a blue stone - which metaphorically depicts the ritual of tempering iron (see in greater detail Ardzinba 1985).
In view of all this the theory of Yakovlev (1941, 140-141), Shagirov (1, 140-141) who tried to explain PAK *ʁʷǝč̣ǝ́ as "dried (ʁʷǝ-) earth (č̣ǝ)" must be obviously taken as a folk etymology. Far more right was Abdokov (1973, 48) who thought the AK and Ubykh words genetically related and explained *ʁʷǝ- as "metal" and *-č̣ǝ as a colour-name. He proposed to consider the second part as "grey", not specifying, however, what particular root he had in mind (thus invoking Shagirov's criticism). Shagirov is right, indeed, that PAK *č̣́ǝxʷǝ 'light-blue, pale' (Ad. č̣́ǝfǝ, Kab. ṣ́ǝxʷ) can not be involved (it has č̣́, not č̣). For Ubykh Abdokov probably meant the root present in Ub. cʷa-bá 'greyish' - which is actually possible (note a quite analogous variation c̣ʷ / cʷ in the Ub. reflexes of PWC *ƛ̣ʷV 'earth' q.v.).
There is more to be said about the hypothetical PWC *ʁ́Iʷǝƛ̣ʷV 'iron'. Since several authors relate the Hatti language to WC (see, e.g., Ivanov 1985), it seems very tempting to relate *ʁ́Iʷǝƛ̣ʷV to the attested Hatti name for 'iron', χap/walki (with χVw- rendering *ʁ́Iʷ- and -lk- rendering the lateral -ƛ̣-), borrowed in Hurrian as χabalgi, and further - Greek khalkó-s, Balto-Slav. *geleǵ- etc. (see Starostin 1985).