As commonly reconstructed, PIE had three different types of "velar-ish" plosives:
- "Palatal velars" (probably plain velar):
- "Plain velars" (probably uvular):
- "Labial velars" (probably labial-something):
We're not sure exactly how they were pronounced; I believe the distinction was probably velar/uvular/labiovelar rather than palatal/velar/labiovelar, but the terminology has stuck, and there's no hard proof either way.
In the centum languages, the "palatal" and "plain" series merged together. For the classic example that gave the phenomenon its name,
*ḱm̥tóm "hundred" > Latin centum, with a
/k/. The "labial" series stayed distinct.
In the satem languages, the "plain" and "labial" series merged together, and the "palatal" series turned into various sibilants, so
*ḱm̥tóm became Avestan satem (and Sanskrit śatam).
Since this root had a "labial"
*kʷ, it merged into
*k in the satem languages rather than becoming a sibilant. That's what you see in Sanskrit. (Slavic languages do seem to have satemized it, as jknappen points out, but that's a later development due to the consonant cluster.)
EDIT TO ADD: As TKR points out, when I say things like "probably uvular", that's very far from certain or uncontroversial. Many linguists think that the sounds were uvular, but many others think they weren't, and there are solid arguments on both sides. Such are the perils of reconstructing!