Meta still pays top dollar

Meta is paying like a company that believes compute is destiny. New reporting based on more than 5,800 of the company’s 2025 work-visa filings shows base salaries for some software engineers reaching $450,000, research engineers hitting $400,000, product managers reaching $348,101, and one vice president of AI listed at $650,000. Most roles still clustered lower, in the $150,000 to $250,000 range, but the outliers are the point. Meta is putting a steep premium on people who can build the systems under the systems. (businessinsider.com, dol.gov) That salary data matters because it is one of the few public windows into how a company values work internally. H-1B and related labor filings do not show stock grants, bonuses, or retention packages. They show base pay only. At Meta, that means the real compensation for top technical hires is almost certainly much higher than the numbers in the filings. The filings also cover a broad swath of the company, including units like WhatsApp and its payments business, which makes the pattern harder to dismiss as a few flashy exceptions. (businessinsider.com, catalog.data.gov, flag.dol.gov) The pattern lines up with what Meta has been doing everywhere else. In its full-year 2025 results, the company said capital expenditures hit $72.22 billion. Mark Zuckerberg used that same earnings report to say he was looking ahead to “personal superintelligence” in 2026. On the earnings call, he said Meta had “rebuilt the foundations” of its AI program in 2025 and was entering a period of “major AI acceleration.” A company does not spend that kind of money on infrastructure unless it also intends to pay for the people who can make the infrastructure useful. (investor.atmeta.com, s21.q4cdn.com) That is why the highest salaries are so revealing. The biggest numbers are not attached to generic coding jobs. They sit around machine learning, systems work, research, and technical leadership. Even when the title is “software engineer,” the market signal is clear: the valuable engineer now is the one who can work close to training infrastructure, model deployment, ranking systems, and the machinery that turns giant models into products for billions of users. Meta’s own AI push made that shift visible last year at LlamaCon, where it rolled out a standalone Meta AI app and a developer-facing API for Llama models. (businessinsider.com, techcrunch.com) The surprise is not that Meta pays well. Silicon Valley has done that for years. The surprise is how narrow the premium has become. Plenty of tech workers are still living through cuts, reorganizations, and stricter hiring. Meta itself has continued trimming teams even as it pours money into AI. That makes the salary bands easier to read. This is not broad generosity. It is targeted spending on a small set of capabilities the company now treats as strategic bottlenecks. One of the clearest details in the filings is also the simplest: about half of Meta’s visa-sponsored roles were for software engineering jobs. (businessinsider.com, cnbc.com)