Bay Area's 7,016-Name WARN Pool: Source Oracle Before Meta

On September 18, 2026, California's EDD reported the Bay Area added 7,800 jobs in August but lost a net 2,900 in tech. Sitting behind that print is a running 2026 Bay Area WARN ledger that names roughly 7,000 engineers across just five employers, each with a legal separation date you can put on a calendar.

If you are sourcing this quarter, this is the cleanest pre-sorted candidate pool you will see all year.

The one-line read on the September 18 EDD print

The Bay Area is rotating talent out of tech and into everything else. August added 7,800 jobs overall while employers cut a net 2,900 tech jobs, meaning the incremental hiring is happening in healthcare, government, and finance while software engineers are the ones being pushed onto the market.

That matters for two reasons most trackers gloss over:

• You are not just competing with other tech firms for these candidates. Non-tech employers hiring in the same metros will move on a Meta or Oracle alum inside a week.
• The 2,900 net is a delta, not a total. Cumulative Bay Area tech layoff announcements through Sept. 18, 2026 sit at roughly 13,900, already 36.6% above the ~10,170 disclosed for all of 2025. The pool is bigger than the monthly print suggests.

13,900

The 2026 Bay Area WARN top five, ranked

Five filers account for 7,016 disclosed Bay Area cuts, roughly half of the 13,900 YTD total. Meta leads on volume, but Oracle is the higher-signal pool by profile density.

The mid-September Oracle WARN alone eliminated 441 Bay Area positions: 279 at the Redwood City HQ, 99 in Santa Clara, and 63 in Pleasanton. That is the single largest one-site tech cut of the month, and it is the one that should be at the top of your sourcing queue this week.

Why Oracle is the sleeper #1, not Meta

Oracle's cuts are the highest-signal pool in the top five because they are homogeneous. The 3,715 Meta number spans Reality Labs, ads, and infrastructure, three very different candidate profiles that require three different pitches. Oracle's 1,093 is concentrated in Redwood City OCI and enterprise apps: a tight cluster of cloud and database engineers whose next employer is almost certainly Snowflake, Databricks, or AWS.

Two facts sharpen the case:

1. Oracle disclosed in its annual report that headcount fell from ~162,000 to ~141,000, a drop of about 21,000 employees over the fiscal year. That is well past the 30,000 March restructuring headline but tells you the alumni flow is already happening, not waiting on the September WARN dates.
2. Oracle raised its 2026 restructuring budget to about $2.8 billion in its September 11 SEC filing. More cuts are funded and coming.

The mechanism: when a workforce shrinks by 21,000 against a 30,000 plan, roughly a third of the announced pool is either still on payroll or leaving on staggered dates over the next two quarters. That is a rolling supply of comparable candidates, not a one-day flood.

Meta is the volume story. Oracle is the profile story. Work Oracle first.

The comp-market anchor nobody is publishing

In Refolk's index of professional profiles, 25,568 US software engineers list Distributed Systems as a core skill. The Meta 3,715 cohort is roughly 14.5% of that entire pool. One WARN filing can move a specialty labor market, and this one already has.

The top current employers of US Distributed Systems SWEs in the index are Databricks, Meta, Datadog, Glean, and Apple. Four of those five are the exact competitive set that will bid for the WARN'd Meta and Oracle alumni. If you are building comp bands for offers this quarter, benchmark against Databricks and Datadog first; they are the price setters.

The site-level filings recruiters should be working this week

September's Bay Area WARN batch is a hunt list, not a headline. Each of these has a physical address and an effective date you can back-time outreach against.

• Oracle Redwood City HQ: 279 separations, largest single-site tech cut of the month.
• Oracle Santa Clara: 99 separations, mostly enterprise apps.
• Oracle Pleasanton: 63 separations.
• Uber: 390 across San Francisco and Sunnyvale.
• PayPal San Jose HQ: 251 layoffs, a compact fintech-engineer pool at one address.
• Intel Santa Clara: 52 separations.
• LeeMah Electronics, Brisbane: 62 separations, a niche EMS/manufacturing pool most tech sourcers will skip.
• Meta Reality Labs, Menlo Park: 272-person filing earlier in 2026, a discrete AR/VR sub-pool with clean crossover to Apple Vision, Snap, and Niantic.

The playbook on each: pull the WARN attachment for job-title breakdowns, pass those titles through a profile search filtered by employer and site, and time your first message to land 30 to 45 days before the effective date. Announced is not executed. WARN separation dates are usually 60+ days out, and the sourcing window opens before the last day on payroll, not after.

This is the exact gap Refolk closes: instead of maintaining four Boolean strings per filing, describe the person in plain English ("Oracle SWE 4 or 5, Redwood City, worked on OCI networking, GitHub activity in 2025 or 2026") and get a ranked shortlist back across GitHub, LinkedIn, and the open web in one pass.

The second five is where response rates actually win

The "second wave" is a better hunting ground than the top five for cold outreach, because the candidates are getting fewer inbound messages per week. Intuit (543), eBay (469), Uber (463), Lucid Group (455), and Walmart (394) round out the 2026 Bay Area top ten.

Why the second five outperforms the first five on response rate:

• Less press coverage per name. Every Meta and Oracle alum has been on a dozen recruiter lists already. eBay alumni have been on far fewer.
• Cleaner profile clusters. Intuit's cuts concentrate in platform engineering. Lucid is embedded, powertrain, and vehicle-software. Walmart Bay Area is Walmart Global Tech, largely ads and search infra.
• Cross-industry crossover. Lucid's 455 is the non-obvious win: EV embedded and systems engineers cross-map cleanly to AI-hardware startups like Cerebras, Groq, and Rain AI, and almost nobody is running that mapping.

If you can only build one sourcing list this week, build it out of the second five. If you can build two, build the Oracle Redwood City list first and the eBay list second.

How to reverse-engineer a WARN filing into a name list

A WARN filing gives you an employer, a site, an effective date, a total headcount, and a job-title breakdown. Convert that into names in four steps:

1. Pull the attachment, not the summary. The California EDD WARN portal publishes PDFs with job-title counts. "Software Engineer 4: 47" is a filterable field. "Layoffs at Oracle" is not.
2. Filter profile data by employer + site + title. Most sourcers stop at employer. Adding site (Redwood City vs Austin) and title bucket (SWE 4 vs SWE 5 vs Principal) cuts the pool to something you can actually work.
3. Cross-reference GitHub and conference talks for sub-team signal. A WARN does not tell you which OCI team an engineer sat on. A recent commit history on oci-python-sdk or a talk at KubeCon does.
4. Score by likely separation date. WARN attachments list effective dates. Prioritize outreach to the earliest cohorts so your message lands before they announce Open to Work and get 300 inbound messages in 48 hours.

The data-hygiene trap: company-wide vs local numbers

Do not quote a company-wide layoff figure and a WARN figure in the same sentence. They measure different things, and mixing them is the fastest way to lose credibility with a hiring manager.

The specific traps in this cycle:

• Amazon announced ~16,000 corporate layoffs in January, bringing total corporate cuts since October 2025 to ~30,000, roughly 10% of its corporate workforce across AWS, Alexa, Prime Video, devices, and advertising. The Bay Area WARN piece of that is 917. Do not tell a hiring manager "30,000 Amazon engineers are on the market in the Bay Area."
• Oracle announced ~30,000 globally in March. The Bay Area WARN piece is 1,093. Actual net headcount drop is 21,000 worldwide across the fiscal year.
• Meta WARN filings total 3,715 in the Bay Area against a much larger announced worldwide plan. The 3,715 is what you can source against with a site and a date.

25,568

What the receiving side of the trade looks like

The same capex reallocation that produced these WARNs is funding the reqs these people should be routed into. U.S. tech has been shifting spending toward AI infrastructure and shedding overlapping roles in software, advertising, and cloud services since 2024, and the pattern accelerated through 2026.

Concretely, the mapping recruiters should run:

• Meta ads and ranking alumni to AI applied-science and ranking roles at OpenAI, Anthropic, xAI, and Perplexity.
• Oracle OCI alumni to Databricks, Snowflake, and AWS cloud-infra teams.
• Amazon AWS and devices alumni to Nvidia, Cerebras, and Groq systems-engineering roles.
• Lucid embedded and vehicle-software alumni to AI-hardware startups needing firmware and low-level systems talent.
• PayPal and Intuit fintech-platform alumni to Stripe, Ramp, Mercury, and Bridge.

The recruiters who win this quarter are the ones who treat the WARN ledger as a sorted destination map, not a doom-scroll feed. The names are already public. The advantage is entirely in how fast you can convert filings into a ranked, contactable shortlist.

FAQ

How current is the 2026 Bay Area WARN data in this article?

The figures are current as of the September 18, 2026 California EDD monthly print and the running 2026 Bay Area WARN ledger published through that date via the state's WARN portal and secondary trackers. The top-five totals (Meta 3,715, Oracle 1,093, Amazon 917, Cisco 706, LinkedIn 585) are cumulative for calendar 2026 to that date, and September batch details (Oracle 441 across three sites, Uber 390, PayPal 251, Intel 52, LeeMah 62) come from filings dated in that same window.

Should I prioritize Meta or Oracle alumni first?

Prioritize Oracle first if you are hiring cloud, database, or distributed-systems engineers, because the 1,093 Oracle pool is far more homogeneous than the 3,715 Meta pool. Meta's cuts span Reality Labs, ads, and infrastructure, which requires three different sourcing pitches; Oracle's are concentrated at Redwood City OCI and enterprise apps and map cleanly onto reqs at Snowflake, Databricks, and AWS.

Why is the second-five list a better hunting ground than the top five?

Intuit, eBay, Uber, Lucid, and Walmart get a fraction of the press coverage that Meta and Oracle do, so their laid-off engineers receive far fewer inbound recruiter messages per week. Response rates on cold outreach to the second-tier list should measurably outperform Meta and Oracle alumni within roughly two weeks of a WARN filing, and the profile clusters are tighter than the top-five headline numbers suggest.

How do I actually convert a WARN filing into candidate names?

Pull the WARN PDF attachment for the job-title breakdown, filter profile data by employer plus site plus title bucket, cross-reference GitHub and conference talks for sub-team signal, and score by likely separation date so you contact the earliest cohorts first. A plain-English search tool collapses the middle three steps into one query across GitHub, LinkedIn, and the open web, which is where most of the time goes when you try to do this by hand.