The war for AI talent just printed its price tag—and it's higher than most CEOs make.
The Summary
- Anthropic's H-1B visa filings reveal two technical staff earning $1.12M and $1.38M in base salary alone—before equity or bonuses in a company now valued at $965 billion
- The "Member of Technical Staff" title is a catch-all that could mean elite AI researcher or seasoned executive making the leap
- Anthropic certified roughly 80 H-1B roles in the first half of fiscal 2026 as it ramps toward a potential $1 trillion IPO this fall
- This is what the talent war for building AGI actually costs, and it's resetting compensation benchmarks across the industry
The Signal
Anthropic just made the invisible visible. Every AI lab pays top dollar for talent, but most keep the numbers locked behind NDAs and equity grants denominated in future valuations. H-1B filings don't lie. They're legal documents, filed with the government, showing exactly what companies promise foreign-born workers in base compensation. And Anthropic is promising some people more than $1 million a year before they even talk about stock.
The context matters. Anthropic hit a $965 billion valuation in May 2026. It's preparing for what could be a $1 trillion IPO this fall. That means early equity grants are worth real money now, not lottery tickets. An engineer who joined two years ago and got even 0.01% of the company is sitting on nearly $10 million in paper value. Add a $1.38 million salary on top and you're looking at compensation packages that rival hedge fund partners.
"The sheer size of these salaries comes from a blazing battle for talent among deep-pocketed competitors like Meta, Google, and OpenAI."
Here's what the numbers tell us about the agent economy:
- The "Member of Technical Staff" title is deliberately vague. It could mean a 24-year-old PhD who published breakthrough research on reasoning models. It could mean a veteran engineer from Google Brain. It could mean someone building the eval infrastructure that keeps Claude from going off the rails. Anthropic keeps it broad so they can pay for impact, not org charts.
- Base salary is only part of the story. These H-1B filings don't show bonuses or equity, which typically dwarf base comp at late-stage startups. If someone's pulling $1.38M in base, their total package likely clears $3-5M annually when you add refreshers and performance bonuses.
- This sets the floor for everyone else. When Anthropic pays seven figures for technical staff, OpenAI and Google DeepMind have to match or lose people. Salary inflation at the top cascades down. Mid-level ML engineers who made $300K three years ago now expect $500K. New grads with strong internships start at $250K.
The timing matters too. Anthropic certified 80 H-1B roles in the first half of fiscal 2026 alone. That's aggressive hiring for a company supposedly preparing to go public. Most startups slow hiring pre-IPO to show discipline and margin expansion. Anthropic is doing the opposite—spending aggressively on talent even as it approaches a trillion-dollar valuation. That tells you something about urgency. They're still in build mode, not harvest mode.
The Implication
If you're building in AI, this data resets your mental model for what talent costs. You're not competing with Google's salary bands from 2023. You're competing with companies that will pay $1 million base to the right person. That means smaller labs need to offer equity that could realistically 10x, or they need to find people who care more about impact than comp. Both exist, but the pool is shrinking.
For workers, the message is clear: specialized AI skills command unprecedented leverage right now. But "right now" is the key phrase. This only lasts as long as the talent shortage does. Once the first wave of foundation models plateau and companies shift from research to deployment, salary inflation will cool. The window for maximizing comp is open, but it won't stay open forever.