AI infrastructure & GPU engineer salaries

These figures come from the salary ranges AI-infrastructure employers publish directly in their own job postings — 75 of 142 current openings disclose a band, and 58 of those are precise enough to use (legal placeholder ranges wider than 2.3× their floor are excluded, the same rule the structured data uses). Unlike a general salary aggregator, every number here is segmented to this one niche: GPU, CUDA, inference, ML-systems and performance engineering, across the AI labs, accelerator makers and GPU-cloud providers actively hiring for it. It is a snapshot of what is open right now, not a survey of the whole market.

The clearest pattern is the gap between frontier labs and infrastructure providers: the labs disclose the highest bands in the set, while GPU-cloud and accelerator companies cluster noticeably lower. By specialty, inference and model-serving roles tend to pay above raw GPU/kernel and performance work. Read the role count (Roles) next to every figure — a median over a handful of postings is indicative, not authoritative, and the numbers shift as roles open and close.