How Amazon Pays

Professional ladder — L4 through L11

Management ladder — L5 through L8

Executive and board

Disclosed executive layer

Track divergence

Hierarchy qualifications and legacy structures

• Amazon publishes no global grade chart, no salary range endpoints and no title-to-level dictionary. Every non-executive row here is a third-party observation or a calibrated model around one.
• L9 does not appear in any public level reference; the ladder moves from L8 straight to L10. Amazon has never confirmed or denied this in a filing.
• The crosswalk is most reliable for software engineering, applied science and product. Sales, operations, advertising and hourly fulfilment roles at the same nominal level can carry a completely different cash and equity mix.
• L1 to L3 frontline and site roles are excluded from this ladder because their pay follows local hourly markets rather than the professional location factors used here. Amazon discloses averages for that population instead: more than $23 an hour in base pay for United States fulfilment and transportation roles and more than $30 an hour including elected benefits, after an investment of more than $1 billion announced in 2025.
• There is no standard Executive Vice President rung. Amazon's public senior nomenclature is Vice President, Senior Vice President, business chief executive and S-team membership, so this report does not invent an EVP band.
• The independent director row is a governance package rather than an employment band, and is excluded from the range chart.

Peer-level mapping

Office and market catalogue — calibration factors versus Seattle

Amazon reports worldwide employees but no reliable office-by-office headcount table, so this catalogue does not rank offices by employee count and every size statement in it is qualitative. Bangalore, Hyderabad, Seattle, Bellevue and Arlington are described as major clusters in the source bundles without a published number attached.

Seattle anchor medians — the basis of every modeled cell

• The Seattle column is the anchor. L4 through L7 medians come from Levels.fyi United States software-engineering records; L8, L10 and L11 are modeled upward from the L7 anchor and the reported director benchmark.
• The management rows use the Levels.fyi Seattle software development manager series for total compensation and model the base, bonus and equity split against the individual contributor mix at the same level.
• Bangalore L5 is a direct observation at 4.75 million rupees base and a 6.21 million rupee package in one bundle and 4.8 million and 6.3 million in the other, so the India anchor is unusually well evidenced by the standards of this report.
• India total compensation runs at roughly 20 percent of the Seattle anchor at L4 and rises towards 61 percent at the top of the modelled ladder, because senior Indian packages carry a far larger equity component while base pay compresses much less.
• Developed markets compress in the opposite direction. London base sits at about 63 percent of Seattle at L4 and rises towards parity at the senior end, while London total compensation falls from about 60 percent to about 52 percent, because Seattle packages become more equity-heavy faster than London packages do.
• The named executive, chief executive and director rows are proxy disclosures converted for comparison only. They are global values, not a local pay range in any selected city, and the location factor applied to them is meaningless.

Executive incentive targets — percent of salary

Named-executive outcomes

• The 222.5 million unvested restricted stock units and the 1.4 billion shares available for future issuance measure different things. Their simple ratio of 15.9 percent is not an Amazon utilisation rate, and no clean plan-utilisation percentage is disclosed anywhere in the filings.
• Treating outstanding units as a share of outstanding units plus available reserve gives 13.7 percent as a capacity ratio. That is an analytical construction, not a reported key performance indicator.
• Unrecognised stock compensation cost was $16.9 billion at the 2025 year-end and had risen to $24.8 billion by 30 June 2026 after the Q2 2026 grant cycle, which is the clearest single signal of how large the annual review grant round is.
• Stock-based compensation expense fell for the second consecutive year, from $24.023 billion in 2023 to $22.011 billion in 2024 and $19.467 billion in 2025, while headcount rose. The 2025 restructuring actions and their $2.7 billion of estimated severance are part of that picture.

Restricted stock unit activity and plan capacity

Vesting — common reported employee schedule

The 5, 15, 40, 40 four-year schedule shown here is the widely reported new-hire offer model and is what public salary databases assume, but it is not what Amazon commits to in a filing. The 2025 Form 10-K and the Q2 2026 Form 10-Q say only that employee awards generally vest over two to five years and primarily vest quarterly in the relevant compensation year. Both statements can be true for different grant cohorts. The practical consequence is that a new-hire package is heavily back-loaded, which is why sign-on cash is normally paid across the first two years, and why leaving in year two forfeits 80 percent of the original grant.

Eligibility by hierarchy level

• Restricted stock units are commonly included for corporate professional roles from L4 upward and appear in public packages at every professional level. Amazon publishes no minimum eligible level, no standard grant table and no guarantee that any individual employee receives a grant.
• There is no evidenced universal corporate restricted stock unit entitlement for L1 to L3 frontline and site roles. Site and hourly programmes differ by facility and country and are not publicly documented.
• Grant timing is the one part of employee equity with real filed visibility: the majority of outstanding awards are granted at hire or in the second quarter as part of the annual compensation review.
• Annualised stock in observed Seattle packages runs about $41,400 at L4, $91,400 at L5, $180,000 at L6, $420,500 at L7 and $919,000 at the director benchmark. At the 25 August 2026 share price of $260.92 those annualised values correspond to illustrative four-year grants of roughly 635, 1,401, 2,760, 6,446 and 14,090 shares respectively. These are model conversions, not disclosed grant targets.
• Unvested units are generally forfeited on departure, including on an ordinary retirement for executives. That forfeiture rule, not any cash retention bonus, is Amazon's principal retention mechanism.

Indicative annual grant value by band — Seattle

Annualized planning value (±25%), not the face value of every new-hire grant. Share equivalents use the $260.92 reference price at 25 August 2026 and ignore plan valuation rules and PSU performance. Highlighted rows are disclosed grant-date stock awards.

Named executive target equity

The single most important thing to understand about Amazon executive pay is that the compensation table and the economics diverge almost completely. Salaries are flat at $365,000 across the layer and carry no differentiation at all. Differentiation lives entirely in the periodic restricted stock unit awards, and in 2025 none were granted. What separates these five executives is the unvested balance each carries: $78.4 million for Herrington, $62.6 million for Garman, $47.9 million each for Olsavsky and Zapolsky, and $242.3 million for Jassy. Those balances are forfeited on an ordinary termination or retirement, which makes them the retention instrument rather than deferred compensation. Jeff Bezos is the outlier in the other direction, holding no unvested equity at all.

Board compensation framework

Because director awards are periodic rather than annual, only four directors show a 2025 grant. The $355,000 figure is the intended annualised value the proxy assigns to a three-year award, not the grant-date value of a single year. Amazon discloses no meeting fees and no cash element at all.

Not publicly disclosed

Modeled promotion planning timeline

The evidence calendar runs from goal and impact evidence in the fourth and first quarters, to organisation-level calibration in the first quarter, to compensation decisions in the first and second quarters, to grants and communication in the second quarter. Only the last step is confirmed by a filing: the 2025 Form 10-K and the Q2 2026 Form 10-Q both state that the majority of outstanding restricted stock units are granted at hire or in the second quarter as part of the annual compensation review.

Dataset summary

City-level H-1B wage history

All-years aggregates. P25/P75 are rounded reconstruction values marked modeled; medians and record counts are the stronger fields.

Selected title / worksite records

Evidence classification

Explicit “Not publicly disclosed” index

Known gaps and diligence before relying on a cell

1. 1The two bundles use different share price and foreign-exchange snapshots. One uses $260.92 at 25 August 2026 with an IMF rate table, the other $262.07 at the 24 August close with a Narodowy Bank Polski table and an INR rate of 95.4075 rather than 95.7143. This report standardises on the 25 August 2026 IMF snapshot and the $260.92 price, which cover every currency used here.
2. 2The bundles disagree on the Seattle anchor by one to three percent at most levels: L4 at $141,000 base and $188,000 total against $138,000 and $183,000, L5 at $179,000 and $274,000 against $177,000 and $271,000, L6 at $215,000 and $397,000 against $214,000 and $406,000, and L7 at $264,000 and $684,500 against $261,000 and $683,000. This report uses the first bundle's figures because they carry the full per-city matrix, and the difference is within the noise of a crowdsourced dataset that moves daily.
3. 3The bundles disagree fundamentally on the L8 row. One models a $330,000 base and $1.08 million total for a senior principal individual contributor; the other reports a $301,222 base and $1.22 million total for a director. This report treats them as different jobs and carries both, as the L8 individual contributor row and the M8 director row.
4. 4City calibration factors are a least-squares line fitted to one bundle's own per-level matrix. Residuals run three to nine percentage points for most markets and reach fourteen points at Tel Aviv, where the reported curve is genuinely non-linear rather than mis-fitted.
5. 5L1 to L3 frontline rows exist in one bundle but are excluded from this ladder, because their location behaviour is the opposite of the professional bands and folding both into one interpolated factor would misstate every developed-market frontline cell by twenty points or more. Amazon's disclosed frontline averages are reported instead.
6. 6H-1B FY2026 volume is irreconcilable between sources at 13,930 applications against 5,201. Snapshot dates, fiscal-year filters and employer-name normalisation all differ. FY2025 volume also varies between 15,529 and 15,524 across snapshots of the same aggregator.
7. 7Amazon publishes no geographic revenue split in either bundle, so this report shows no regional revenue panel rather than substituting the reportable segment structure for one.
8. 8City-level H-1B petition counts are not exposed in either cited employer aggregate. The record count column is recorded as zero and the percentiles are explicitly modelled.
9. 9The peer chief executive comparison differs between bundles. One uses Alphabet, Meta, eBay and Microsoft; the other uses Microsoft, Alphabet, Walmart and Apple. This report carries the five with the clearest fiscal-year attribution and notes eBay's Jamie Iannone at $28.46 million as the alternative retail peer.
10. 10No company-wide 2026 salary hike percentage, pay freeze, salary cut, campus offer revision or off-cycle correction was located in any source. The 2025 pay-model changes described in press reporting are not an Amazon disclosure.

FX rates used — 1 USD equals, snapshot 2026-08-25

Single-date conversion layer for comparability; it ignores payroll-date FX, tax, purchasing power, benefits valuation and hedging. Compensation intelligence, not legal, tax, investment, immigration or employment advice.

Source register — 17 sources

Total headcount kept rising through both restructuring rounds because the cuts were concentrated in corporate roles while the operations network kept growing. Amazon publishes no split between corporate and operations headcount, no city-level or country-level table and no attrition rate, so the only reliable geographic signal in the filings is which markets have offices at all. Seasonal and contractor treatment also differs between reported years.