Amazon AGI Salary (2026)

Two Orgs, Two Pay Curves

Amazon's AI work splits cleanly. Amazon AGI is the model-building org. It owns the Olympus training program, the Nova model family launched at re:Invent 2024, and the research staff who report up through Rohit Prasad. AWS AI services sits inside the cloud business and ships products: Bedrock as the model gateway, SageMaker for training infrastructure, plus Q, Comprehend, and the rest of the managed AI lineup.

Pay reflects that split. Amazon AGI compensation now tracks frontier labs more than classic Amazon. Senior Applied Scientists and Members of Technical Staff inside the Olympus and Nova teams sit at L6 or L7 with total comp that often clears $500K, sometimes $700K, when the equity refresh stacks. AWS AI services pays closer to the historical AWS curve. Strong but not frontier-lab money. Senior SDE roles cluster around $300K to $400K total comp.

Why the gap? Retention. Amazon AGI competes for the same researchers that OpenAI, Anthropic, and Google DeepMind want. The Anthropic partnership, announced in stages across 2023 and 2024 and now totaling $8B in Amazon investment, sits inside the same building from a comp perspective. Amazon's internal AGI staff watch Anthropic salaries and ask for parity. AWS AI services hires a different profile, often strong applied ML engineers without the same outside-offer leverage.

Amazon Levels and Total Comp

Amazon's level structure runs L3 through L12. AI engineers and applied scientists live mostly in the L4 to L8 range. Numbers below come from Levels.fyi public submissions for 2025 and early 2026, weighted toward the Amazon AGI and AWS AI orgs where they can be isolated.

Amazon's base ceiling matters. For years the company capped base salary near $185K and made up the rest with stock, paid heavily in year 1 sign-on and year 2 sign-on to compensate for back-loaded RSU vesting (5% year 1, 15% year 2, 40% year 3, 40% year 4). That cap loosened around 2022 when Amazon raised the maximum base toward $350K for SDEs to compete with Google and Meta. Inside Amazon AGI specifically, several L7 packages confirmed on Levels.fyi show base salaries above $300K, which would have been impossible three years ago.

Amazon AGI vs Bedrock vs Frontier Labs

The hard comparison is Amazon AGI senior comp against Anthropic. Anthropic packages senior research engineers at $700K to $1M+ with equity that has appreciated as the company moved from $18B to a reported $60B+ valuation across late 2024 and 2025. Amazon AGI L7 senior staff sit closer to $500K to $750K, with RSUs tied to a stock that grew but did not 4x like Nvidia. The retention gap is real and the team manages it case by case with special equity grants and accelerated promotions.

Inside Bedrock the picture is different. Engineers building the Bedrock control plane, the inference routing layer, and the customer-facing API are SDE roles at AWS rates. They cap out lower than AGI research but ship to revenue. An L6 senior SDE on the Bedrock team makes $300K to $420K total comp. That is excellent money but not frontier-lab money. The work is also more applied. Engineers report shipping to customers daily versus writing training code that may or may not ship.

Compare to Nvidia, where stock appreciation has turned mid-level engineers into millionaires through vesting alone, or to Netflix, where the all-cash philosophy puts $700K to $1M base on the table for senior ML staff. Amazon AGI competes hardest on mission and access to compute. The Olympus training runs use clusters that few companies can match.

The Anthropic Effect

Amazon's $8B investment in Anthropic is the largest single factor reshaping internal AI comp. The deal gives Amazon access to Claude on AWS Bedrock and gives Anthropic a major customer for compute via AWS Trainium chips. Internally it does something else. It puts a price tag on outside-offer risk.

Amazon AGI staff who get an Anthropic offer can show it to their manager. The numbers are usually 1.5x to 2x their current Amazon package. Managers can match in roughly two ways: a one-time retention RSU grant on top of the standard refresh, or an accelerated promotion that bumps the next level's stock budget. The retention RSU path has become more common since late 2024. Levels.fyi submissions tagged as "retention" or "off-cycle" inside Amazon AGI now show grants of $500K to $1.5M vested over two to four years.

The same pressure does not exist inside AWS AI services. A Bedrock engineer rarely gets an OpenAI offer. The skill profile is closer to systems engineering than research. That keeps AWS AI comp tracking the broader AWS curve, which moves slower than the Amazon AGI curve.

Culture Tax and PIP Risk

Amazon's leadership principles cut both ways. The system rewards bias for action and customer obsession. It also rewards visible delivery, which inside AGI can mean shipping demos and papers on a quarterly drumbeat. Several public posts from former AGI staff describe pressure to ship results that map cleanly to Andy Jassy's quarterly AI narrative. That pressure is the cultural tax on the high comp.

The other dynamic is unregretted attrition, internally called URA. Amazon manages a target attrition rate by level. Managers who do not hit the URA target face pressure to performance-manage out lower performers. For an L6 with strong reviews this is a non-issue. For an L5 borderline performer it is real. The phenomenon is well documented across the company and applies inside Amazon AGI as well. New hires negotiating offers should factor that risk against the year-2 sign-on cliff, which is the most painful point if you leave between years 2 and 3.

Signing Bonuses and Year-2 Cliff

Amazon's compensation structure is back-loaded. A typical L6 offer in Amazon AGI looks like this: $250K base, $180K year-1 sign-on, $120K year-2 sign-on, $400K RSU vest split 5/15/40/40 across four years. Total year-1 comp lands near $430K and year-2 near $430K. The cliff arrives in year 3 when the sign-on is gone and the RSU vest jumps. If the stock holds, year-3 and year-4 comp jumps to $400K plus. If the stock drops, the engineer ends up underwater and starts taking calls.

Negotiate the sign-on aggressively. Amazon recruiters expect counter-offers and have visible flexibility on the sign-on numbers, less on base, less still on the RSU grant. Outside competing offers help most. The strongest leverage right now is an Anthropic or OpenAI offer in hand. Inside AGI, recruiters will move quickly to match those.

What This Means for AI Leaders

For Chief AI Officers and CTOs hiring against Amazon, the playbook is clear. Amazon AGI is a strong recruiter but loses on cash compensation to Netflix, on equity upside to Nvidia, and on mission focus to Anthropic. Sell on those gaps. If you cannot match the cash, sell on scope, ship velocity, or the chance to own a product end to end. Amazon AGI engineers often describe working on small slices of large training runs. That is unattractive to a senior engineer who wants to architect a system.

For candidates, the question is what you optimize for. If you want to do frontier research at scale on huge clusters and accept that you are inside a large org, Amazon AGI is competitive. If you want maximum cash, Netflix beats it. If you want maximum equity upside on a smaller team, Anthropic or a Series D startup beats it. If you want stable compounding on a high-quality public stock, Nvidia still beats it through pure appreciation.

Related Salary Guides

Compensation data for related companies and roles:

• AI Company Salaries (2026). Master hub covering 30+ frontier and enterprise AI companies.
• Nvidia Salary (2026). The other AI infrastructure giant, paid largely in appreciated stock.
• Netflix Salary (2026). The all-cash counter-model to Amazon's back-loaded RSU structure.
• AI Lab Levels Explained. How L-levels, MTS, and Senior Staff translate across companies.
• AI Lab Millionaires. Where stock appreciation has done the heavy lifting.
• AI Engineer Salary (2026). Cross-industry compensation guide for the role.