Amazon's 2,303-Person WARN Has 9 Dates: The 6-Month Calendar

Amazon's Oct 28, 2025 Washington WARN filing is not a layoff. It is nine layoffs, staggered across five months, with a follow-on filing that pushes the runway into late June 2026. If your sourcing plan is "block a Tuesday and blast the list," you will miss most of the 2,303 people and all of the ones worth hiring.

The filing lists roughly 2,303 employees across 56 Washington facilities, with separations effective on nine specific dates: January 26, January 30, February 5, February 25, February 27, March 19, April 1, April 26, and May 26, 2026. A second WARN dated January 29, 2026 added 84 more roles across 30-plus Seattle and Bellevue offices, separating between February 2 and February 23. That is a rolling six-month sourcing window with at least a dozen named cohorts inside it.

Why Amazon WARN 2026 is a calendar, not an event

Amazon is giving at least 90 days' notice (30 more than WARN requires), and per employee accounts on Blind, badge and network access get revoked at notification, not at the separation date. Every "separation date" on the PDF is actually the end of a paid-but-locked-out window that began 60 to 90 days earlier.

For sourcers, that means the useful anchor is not the separation date. It is notification plus roughly seven days, which is when the person has processed the news, told their spouse, and started answering messages. If you wait for the WARN date on the PDF, you are hitting people who already took an offer 45 days ago.

Amazon HR chief Beth Galetti framed the cuts as being "organized more leanly, with fewer layers and more ownership." CEO Andy Jassy was blunter on the Q3 2025 earnings call: "The announcement that we made a few days ago was not really financially driven, and it's not even really AI-driven, not right now at least. It's culture." That framing matters. These are not obsolete skills being retired. They are performing engineers, PMs, and designers caught in a span-of-control cleanup.

2,303

The nine-date wave calendar

Build four waves, anchor each one to the notification window (roughly 90 days before the listed separation), and run outreach from 30 days before separation through 60 days after. Below is the calendar derived from the two Washington filings.

Two things to notice. First, waves overlap. Somebody separating May 26 was notified in late February, which is inside the outreach window for Wave 2. If you run one campaign per wave without deduping against the previous one, you will send the same person three "sorry about Amazon" emails from three of your recruiters.

Second, the outreach window is not symmetric. The pre-separation weeks are golden because the candidate is still on payroll, still has severance math on their mind, and is comparing offers with time pressure. The post-separation weeks are noisier because everyone else finally caught up. Front-load your calendar.

The Seattle pool is bigger than the WARN

In Refolk's index, roughly 15,091 US-based profiles currently or recently list Amazon or AWS with an SDE-family title, and Seattle is the densest single region, ahead of Hyderabad and Bengaluru. The 2,303 Washington names on the WARN are a fresh, high-signal slice of a much larger addressable pool.

Do the math. If even 5% of the 2,303 Washington separations are SDE-title and Seattle-located, that is roughly 115 SDEs entering the market, or about 4% of the current addressable Seattle Amazon SDE pool visible in Refolk's index in a single quarter. That is a real dent in a market that was already the tightest in the country, and it is why every Series B engineering leader in Pioneer Square is refreshing the WA ESD portal on Tuesday mornings.

The "SDE" title itself is worth calling out. Refolk's index shows that a search on "Software Development Engineer" (as opposed to the generic "Software Engineer") returns a pool where Amazon and AWS together account for more than 60% of the top-company distribution. SDE is not a job title, it is a fingerprint. Split your Boolean into SDE, SDE II, SDE III, Sr. SDE, and Principal Engineer buckets rather than one blob, because each bucket has a different comp band and a different willingness to leave Seattle.

15,091

Sourcing laid off Amazon employees without the swarm

Skip the engineers on day one, work the non-engineering cohort first, and set up a weekly WA ESD scrape rather than a monthly one so you catch the mini-filings before your competitors do.

The Jan 29 follow-on filing is the tell. It lists 84 roles including software development engineers, program managers, recruiters, HR specialists, and UX designers, ranging from entry-level to directors and principals. Washington's WARN regime tightened in July 2025 with new disclosure requirements for terminations within 90 days of a prior notice. That is why these small cleanup cuts are suddenly visible. Before July 2025, Amazon could have folded the 84 into the 2,303 or not disclosed them at all. Now it cannot, and you should expect three to six more mini-filings through mid-2026, each a fresh named list.

Here is where competition is thinnest:

• RecOps and People Analytics. The Jan 29 filing explicitly names HR specialists and recruiters. Nobody's engineering VP wakes up wanting to hire ex-Amazon recruiters, which means less inbox noise for these candidates.
• UX and Content Design. Named in the same filing. Design cohorts inside Amazon are small and tightly networked, so a warm intro through one lands three more.
• Program managers (non-technical). Amazon's TPM population gets swarmed. Non-technical PMs and Chief-of-Staff types are undervalued and often the best operators in the org.
• Middle managers. Jassy said "fewer layers." Managers of 2 to 8 who were performing well are collateral damage, and they read differently on a resume than they play in an interview.

For the SDE cohort, the swarm is real. You need a different edge, which is where describing the person in plain English beats a Boolean string. Refolk lets you say "SDE II or SDE III in Seattle who shipped services on AWS in the last two years and has a GitHub with recent commits," and get a ranked shortlist back. That is the exact gap keyword search leaves open when 400 recruiters are all running the same "Amazon" + "Seattle" filter on the same Tuesday.

The 56 facilities give you geo-fences most sourcers miss

The Oct 28 PDF names 56 specific Washington facilities, including the Bellevue SEA93 tower and DSE8 in Bremerton, and each facility code corresponds to a team cluster you can target with location-scoped LinkedIn and GitHub searches.

Facility codes are not just addresses. Inside Amazon they map to org clusters. SEA93 is not the same team as the Bremerton DSE8 population. If you have the facility list from the WARN PDF, you can:

1. Cross-reference facility codes against team blogs, GitHub org membership, and public re:Invent talks to guess which product areas got hit.
2. Set LinkedIn geo-filters at the sub-city level (Bremerton, Redmond, Bellevue East) rather than the metro level.
3. Prioritize candidates whose LinkedIn "current location" matches a facility that appears on the WARN, because those are the people who cannot easily commute to a still-standing Amazon office.

This is grunt work, and it is worth it. The recruiters running "Amazon" + "Seattle, WA" in LinkedIn Recruiter are competing with each other. The recruiter who searches "Amazon" + "Bremerton" is competing with almost nobody.

The layoff date on the PDF is the end of the sourcing window, not the start of it.

What history says about the follow-ons

Warntracker.com shows Amazon has filed 205 WARN notices for a cumulative 30,890 workers from February 2001 through September 2026, and the recent pattern is small cleanup filings every 30 to 60 days after a big one. Plan for them.

Concrete recent evidence: a July 2 filing hit 57 employees by August 3. An October filing added 121 more, with the first employee scheduled to be laid off on October 1, thirty-two by October 2, 66 by October 20, and the last separation on October 27. That is the pattern to expect through summer 2026 in and around Seattle.

The broader context: Amazon announced roughly 16,000 corporate cuts on Jan 28, 2026, its second round of mass job cuts since October, and cumulatively the 30,000 cuts represent about 10% of a corporate and tech workforce of roughly 350,000. Washington is a disproportionate share of that. Your calendar should assume new WARN filings land on a two-week cadence, not a quarterly one.

A practical weekly cadence for the next six months

Run four workflows on a weekly cycle: scrape, enrich, prioritize, and reach out. Below is what each looks like in practice.

• Monday: scrape. Pull the WA ESD WARN portal (fortress.wa.gov/esd) and the GeekWire PDF host (cdn.geekwire.com). Diff against last week. Anything new goes into a fresh sheet with the facility code, separation date, and estimated notification date.
• Tuesday: enrich. Take the previous week's names and enrich against LinkedIn, GitHub, and the open web. This is where a plain-English tool matters, because "SDE III at Amazon who shipped a public package on GitHub in the last 90 days" is a description, not a filter. Refolk turns that description into a ranked list without you writing a Boolean.
• Wednesday: prioritize. Sort by (a) proximity to notification date, (b) match to your open req, (c) whether they show up in your ATS already. The people 7 to 30 days post-notification are your hottest cohort.
• Thursday and Friday: reach out. Personalized first message referencing something specific. Not "sorry about Amazon." Something like "your talk at re:Invent 2024 on X" or "your commits to Y." Then a two-message follow-up cadence over the next 14 days.

If you are staffing a 50-person engineering team out of this cohort, you will run this cycle for 26 weeks. If you are a single sourcer supporting one founder, you will run it for maybe 8. Either way, the calendar is the deliverable, not the list.

FAQ

How do I access the Amazon WARN filing directly?

Two primary sources. The Washington Employment Security Department maintains the WARN portal at fortress.wa.gov/esd, which is the canonical database and updates within a few days of filing. GeekWire hosts a PDF of the Oct 28, 2025 Amazon filing at cdn.geekwire.com, which is the fastest way to see the full 56-facility, 9-date breakdown without paginating through the state UI. Bookmark both, because the state portal is authoritative but the news outlets often post the PDF first.

When are laid off Amazon employees actually on the market?

Earlier than the WARN separation date suggests. Per employee accounts, Amazon typically revokes badge and network access at notification, then keeps employees on payroll for the 90-day notice period. That means somebody with a listed separation date of May 26, 2026 was notified around late February and has been functionally job-searching since March. Anchor your outreach to notification plus roughly seven days, not to the PDF date.

Is this an AI-driven layoff, and does that change how I pitch?

No, and yes. Andy Jassy explicitly called the cuts a culture and layers story, not an AI-replacement story, and Beth Galetti's public framing matched. That matters for your pitch because affected employees were not cut for obsolete skills. Do not write outreach that implies they need to reskill or that positions your role as a soft landing. Write it as a lateral or step-up move for a performing engineer whose org just got flattened.

What about the non-Washington cuts?

The 2,303 and 84 figures are Washington-only, because WARN is state-level. Amazon's broader Jan 28, 2026 announcement of roughly 16,000 corporate cuts includes California (Amazon Games in Irvine and San Diego), plus offices in other states. Those cohorts will surface in state WARN databases on their own timelines. Set the same weekly scrape for whichever states matter to your reqs, because none of them will publish on Amazon's schedule.