Went through the NVIDIA Staff SWE process at the end of last year and got an offer I ended up turning down (took something else, mostly about the team). Posting the numbers because I couldn't find anything accurate at this level.
Staff SWE (NVIDIA's internal equivalent, ~10-14 YOE, Santa Clara, early 2026): Base: $250k - $280k RSUs: $800k - $1.4M over 4 years Bonus: 20-25% target (and NVIDIA has actually paid these the past several years) Year-1 TC: somewhere in the $460k - $580k range depending on grant size and stock price at vest
My specific offer: $265k base, $1.1M RSUs over 4 years (cliff at 1 year, then quarterly), 22% bonus target. On paper about $545k year-1.
On the equity structure: it's 4-year vest with a 1-year cliff, then quarterly. Pretty standard. Grant date price matters a lot. They refresh equity annually but it's not guaranteed and the amount varies with performance and budget.
Why I turned it down: combination of strict in-office expectation (5 days for my team), and the specific charter of the role was narrower than what I wanted. The money was genuinely impressive.
The Staff-level loop at NVIDIA: Two system design rounds (mine were GPU memory management and distributed training infrastructure at scale) Two coding rounds (hard-ish LC, but they moved fast, not 45 minutes on one problem) One leadership/influence round: how do you get things done without authority, cross-team alignment, a project where you shaped the technical direction One executive conversation (VP level, more of a fit conversation than a grilling)
The system design questions at staff level assume you can talk about GPU-specific tradeoffs. I had 8 years of ML infra experience so this was fine for me, but if you're Staff-level from pure web infra you need to spend real time getting up to speed on GPU memory, NVLink, distributed training patterns before going in.