I run the technical and operations side of TTGC, which means I see real comp data across roles every quarter — what we pay, what our competitors pay, what candidates come in asking for. Let me cut through the hype.

The headlines say AI engineers make $1M. That's true for maybe 100 people at OpenAI and Anthropic. For everyone else, the comp picture is more nuanced — and more interesting.

The 2025 compensation picture

Multiple independent compensation surveys ran in 2024-2025 and reached similar conclusions. Levels.fyi (which aggregates real comp data from offer letters and signed compensation) shows median total comp for AI/ML engineers at top tech firms ranged from $250K-$450K in 2024. Robert Half's 2024 Salary Guide reported similar ranges for the U.S. broader market.

But the top of the market and the median have diverged sharply. Here's where the money actually is in 2025.

Tier 1: Foundation model researchers ($500K-$5M+)

These are the researchers building the models themselves at OpenAI, Anthropic, Google DeepMind, xAI, Meta's AI lab, and a handful of others. New PhD researchers regularly receive offers exceeding $500K total comp. Senior researchers exceed $1M routinely. The very top — researchers leading critical capability areas — have reportedly received offers in the $2M-$10M range, per multiple 2024 reports from The Information and The Wall Street Journal.

Realism check: this market is real but tiny. We're talking about a few hundred jobs globally. PhD plus published research is effectively required.

Tier 2: AI infrastructure engineers ($300K-$700K)

The people who build and run the infrastructure that AI models train on — CUDA programmers, distributed systems engineers, GPU cluster operators, ML platform engineers. Less famous than the researchers, often paid almost as well, because the supply is even more constrained. Levels.fyi data from 2024 shows senior infrastructure engineers at top labs in the $400K-$700K range.

These roles typically require a strong CS background plus distributed systems experience. They're hard to enter from the outside, but easier than research.

Tier 3: Senior ML engineers at well-funded startups ($250K-$500K)

The applied side. Engineers who can take a foundation model and turn it into a production product. Companies like Perplexity, Cursor, Glean, Harvey, and dozens of other AI-native startups compete for this talent. Pay is high but more volatile — heavily weighted toward equity that may or may not be worth what the cap table suggests.

Required skills: strong engineering, comfort with rapid iteration, ability to ship products with AI in the loop. Not necessarily an ML or CS degree, but you need a portfolio of shipped AI products.

Tier 4: AI engineering at enterprise companies ($150K-$280K)

Banks, insurance companies, healthcare systems, retailers — all of them are building internal AI teams. The pay is lower than the top labs, but the work is interesting and the comp is real. Robert Half's 2024 Salary Guide reports a median of $180K-$220K for senior ML engineers at U.S. enterprise companies.

This is the most accessible high-paying AI engineering market. You don't need to be in San Francisco. The roles are distributed across all major U.S. business hubs.

Tier 5: AI-adjacent non-engineering ($80K-$200K)

AI product managers, AI program managers, AI policy specialists, AI sales engineers, AI consultants. These roles benefit from the AI economy without requiring deep engineering chops. Pay is solid but well below the engineering tracks. Indeed Hiring Lab data from 2024 shows median total comp in the $100K-$160K range for these roles in the U.S.

These roles are some of the fastest-growing in absolute number of hires. They're where the broader AI economy actually employs people.

Tier 6: Prompt engineers and AI content specialists ($60K-$130K)

These were briefly hyped as the next $300K job. The reality is more modest. Indeed and Levels.fyi data from 2024 show prompt engineer median salaries in the $80K-$130K range — solid pay but not the unicorn comp the early hype suggested.

The reason: prompt engineering as a standalone job is being absorbed back into adjacent roles. Content strategists, product managers, and ops folks are picking up prompt skills, and the pure-play prompt engineer role is consolidating.

Where the money is actually moving

Comp data from 2023-2025 shows three clear trends:

What pays well that nobody talks about

Three under-discussed high-comp tracks:

The honest comp advice

If you want to maximize total comp, the path is: pursue a PhD in ML, publish, get hired into foundation model research. Tiny field, enormous comp.

If you want a strong AI career with realistic odds, pursue: strong engineering fundamentals + AI specialization + ship products that demonstrate the work. Then optimize your geography and company stage based on whether you prefer cash or equity.

If you don't want to be an engineer at all, the non-engineering AI roles pay well — not lottery-ticket well, but well above median professional comp.

In all three cases, the salary follows real skills and shipped work. The market in 2025 is honest in a way it sometimes hasn't been: people who can do the work get paid. People who can't, don't.

Sources