What Meta’s visa filings tell HR leaders about the real cost of AI talent
Meta's visa filings showing $650K base salaries aren't a benchmark — they're a warning shot. Most companies aren't competing for that talent, and their workforce planning tools aren't built to help them figure out what they're actually competing for.
The news
HR Executive reported on what Meta’s H-1B visa filings reveal about AI talent costs, with base salaries reaching as high as $650,000. The piece cautions HR leaders against using these figures as direct benchmarks — but acknowledges the data is already circulating and influencing conversations.
My take
The instinct to benchmark against Meta’s AI salaries is understandable. When a number like $650K lands in the news cycle, it becomes a reference point whether you want it to or not — in board meetings, in candidate negotiations, in “why is our offer getting rejected” conversations with hiring managers.
But here’s what bothers me about how most companies will respond to this: they’ll either panic-adjust or dismiss it entirely. Both reactions are wrong, and both stem from the same underlying problem — most organizations don’t have workforce planning infrastructure sophisticated enough to tell them which answer is right for their situation.
The real question isn’t “how do we compete with Meta’s AI salaries?” It’s “which AI roles in our organization are actually competing in that talent pool, and which ones aren’t?” A computer vision researcher building foundation models is not the same as a data scientist standing up an AI-assisted HR workflow. They don’t compete in the same market. They don’t command the same compensation. But I’ve seen companies treat them identically — and get burned both ways, either overpaying for roles that didn’t require frontier talent or losing genuine research capabilities because they anchored to the wrong benchmark.
Workforce planning tools should be helping HR leaders draw those distinctions. The vendors building AI-powered compensation and planning products — Compa, Syndio, Visier, Lightcast — have the data infrastructure to segment this. The question is whether HR teams are actually using those tools at this level of precision, or whether they’re still exporting spreadsheets and guessing.
I suspect it’s mostly still the latter. And if that’s true, a $650K headline is just noise.
The so-what
I’d tell my clients: before this number shows up in a board deck or a recruiter debrief, get clear on exactly which roles in your AI hiring plan are playing in the hyperscaler talent market and which ones aren’t. That segmentation is a workforce planning problem, not a compensation problem — and it should be solved before you’re sitting across from a candidate.
The companies that will win on AI talent aren’t the ones who match Meta’s salaries. They’re the ones who know precisely when they need to and when they don’t.