Cloudflare's 1,100 Get Paid Through December. Source Them in September.
Cloudflare's May 7 cut of 1,100 has full pay through Dec 31. Here's why Q4 is too late and how to source this AI-native cohort now.
On May 7, 2026, Cloudflare announced its first mass layoff in 16 years: 1,100 people, about 20% of staff, gone in a single restructuring framed as a pivot to an "agentic AI-first" operating model. The instinct for most recruiters is to fire up LinkedIn alerts and chase the cohort over the next two weeks. That instinct is wrong this time. Cloudflare is paying full base salary through December 31, and the actual sourcing window opens in late summer, not next month and not in Q4.
Why the standard layoff playbook breaks here
Every layoff playbook you've read in the last three years tells you the same thing: hit the cohort in the first two to three weeks, before they go cold or get scooped. That advice was calibrated on cohorts like Oracle's March 31 cut of roughly 30,000 (12 to 16 weeks of severance, most senior engineers placed inside 17 to 30 days) or Meta's May 20 cut of about 8,000 (16 weeks plus two extra weeks per year of service). Those people are financially forced to move fast.
Cloudflare's package is a different animal. Departing employees keep full base salary through December 31, 2026, US healthcare extends through year-end, and equity vesting is accelerated through August 15 with one-year cliffs waived in favor of pro-rated vesting. Do the math against Meta. A Meta engineer with four years of tenure gets 24 weeks of pay and hits the market hungry by late August. A Cloudflare engineer cut on May 7 has roughly eight months of full pay. That is four times the runway of the Oracle cohort and roughly double the Meta cohort.
Runway changes behavior. Cohorts with two months of severance take the first reasonable offer. Cohorts with eight months of severance run multiple processes in parallel, talk to friends at Anthropic and Stripe first, and ignore cold InMails entirely until they're ready. They are picky, patient, and they negotiate. If you treat them like a distressed pool, you will either annoy them or lowball them.
The cohort is not who you think
Matthew Prince's WSJ op-ed three weeks after the announcement said the quiet part out loud: "the vast majority of those we laid off last week were measurers." He defined measurers as middle management, finance, legal, internal audit, and revenue recognition. Builders, meaning engineers, and revenue-carrying sales reps were largely retained.
Read that carefully before you assume "Cloudflare layoffs 2026" means a flood of senior infrastructure engineers. It mostly doesn't. But the engineering slice that did get cut is the most interesting part of the 1,100. Those are engineers whose function (internal tools, platform glue, ops engineering, data plumbing) was absorbed by agents inside Cloudflare. Prince told earnings that Cloudflare's AI usage grew more than 600% in the last three months alone, with thousands of agent sessions running daily across engineering, HR, finance, and marketing.
In other words: the engineering subset of the 1,100 didn't get cut because they can't code. They got cut because they spent the last year being supervised by, or supervising, agentic systems that ate their job. That is the deepest possible training signal for what the market keeps calling "AI-native engineers."
They didn't get laid off because they can't code. They got laid off because they watched agents eat their function from the inside. </pull> If you're hiring agentic AI engineers, this is the cohort that has actually lived it, not the one that put "agentic" on their LinkedIn headline last Tuesday. ## The two cliffs that actually matter Severance ends December 31. That date is a red herring for sourcing. The real inflection is August 15. August 15 is when equity vesting stops accruing. That is the emotional cliff, not the financial one. Engineers who have been telling themselves "I'm still kind of at Cloudflare, the equity is still vesting" stop telling themselves that on August 16. The identity shift happens there. That's when "I should probably start looking" turns into "I'm running a real process." The second cliff, December 31, is when every recruiter in North America hits this cohort simultaneously. By November, you will be competing with Anthropic, OpenAI, Stripe, Anduril, Ramp, Brex, and roughly forty AI-infra Series B companies for the same 1,100 names. Inbound to these people in Q4 will be unreadable. Their inboxes will look like an Oracle alum's inbox did in April. The arbitrage window is the eight to twelve weeks between mid-August and early November. Conversations, not offers. Coffee, not pipeline. ## How to actually source them now Three things to do between now and Labor Day, none of which involve sending a cold pitch about your Series B. ### 1. Build the list before competitors do Most ATSes will surface "ex-Cloudflare" in November when LinkedIn headlines update. By then you're late. The list exists now, it's just not labeled. You want to identify the engineering subset of the 1,100, filter for the agentic and platform functions that were specifically eaten by agents, and weight for the geographies most US recruiters miss. Per our index of Cloudflare engineers and EMs, the bulk concentrate in Austin, the SF Bay Area, the UK, and Lisbon, with a long tail in India and Germany. Austin and Lisbon are systematically underweighted by US sourcing teams. This is the kind of compound query that breaks Boolean. You want "engineers who were at Cloudflare on May 6, are no longer listed there as of June, worked on internal tooling or platform infra, and currently live in Austin or Lisbon." That's a plain-English question, which is why we built [Refolk](/): you describe the person in plain English and get a ranked shortlist across GitHub, LinkedIn, and the open web instead of stacking forty Boolean operators that LinkedIn will silently truncate. ### 2. Lead with the work, not the company This cohort will smell AI-washing from a mile away. They watched their own employer go from "we use AI internally" to "AI replaced 1,100 jobs" inside twelve months. They know the difference between agentic systems in production and a press release. Job descriptions and outbound that lean on "agentic," "AI-first," and "10x productivity" without naming a specific stack (which model, which orchestrator, which eval harness, what the human-in-the-loop pattern looks like, what's in production versus pilot) will get filtered before you finish the InMail. If you can't describe the agentic system the new hire would supervise in three concrete sentences, don't send the message yet.
refolk prompt: Engineers who left Cloudflare between May and July 2026, worked on internal platform or agentic tooling, and are based in Austin or Lisbon. note: You get a ranked shortlist with GitHub activity, recent commits, and the specific Cloudflare team they were on, not a CSV of every ex-Cloudflare engineer on LinkedIn. slug: 91esner2rm
### 3. Run the September conversation, not the November pitch
The recruiters who will win this cohort are the ones in their inbox in September with a low-pressure intro: "Saw you left Cloudflare. We're building X with Y stack. Not pitching a role, just wanted to be on your radar when you start looking in November." That message lands when their equity cliff just hit and their calendar is empty. The same message in November lands at the bottom of a stack of forty.
Founders who try to compress this into a single November offer cycle will lose to founders who started the relationship in August. There is no shortcut. Sourcing laid off engineers from a patient cohort is a relationship game played on the cohort's timeline, not yours.
## What this cohort signals about the broader market
A Challenger, Gray & Christmas report counted 49,135 US layoffs tied to AI so far in 2026, nearly matching the entire 2025 total. Cloudflare is the cleanest signal in that stack because Prince said the quiet part on the record and because the stock got punished for it. Shares dropped more than 20% on May 7 despite Q1 revenue hitting $639.8M, up 34% year over year and the company's best quarter ever. Wall Street isn't buying the "AI productivity replaces headcount" thesis yet, but the talent market has to plan as if it's real.
Prince also said Cloudflare will keep hiring and that 2027 headcount will likely exceed any point in 2026. The engineers being retained are the forward template for what the market wants next: people who can design, supervise, and audit agentic systems in production. The 1,100 includes the closest available proxy for that profile outside the company's current walls. That's the trade.
The companies that will quietly absorb the best of this cohort (probably Anthropic, probably a handful of AI-infra Series Bs nobody is watching) will do it in September and October. By the time the WARN-style trackers and "Open to Work" badges catch up, the top quartile will be off the market. If you're sourcing against a Q4 plan, move it up. Use whatever stack lets you build the list now from a plain-English description (Refolk does this, but the principle holds with any tool that can search across GitHub commits and the open web, not just LinkedIn headlines). Then sit on the list, warm it through August, and run the real process in September.
The Cloudflare severance package wasn't designed to help you hire faster. It was designed to give departing employees dignity and time. That same time is the gap a smart sourcer plays in.
FAQ
When will the Cloudflare 1,100 actually start interviewing?
Most will start serious conversations in mid-August, when equity vesting ends, and real processes in September and October. A small minority who never planned to take a break will be active in June, but the bulk of the cohort is using the runway intentionally. Treating December 31 as the start of the window is too late; by then you're competing with every recruiter in tech for the same names.
How is this different from sourcing the Oracle or Meta layoff cohorts?
Runway and composition. Oracle alumni had 12 to 16 weeks of severance and placed inside a month; Meta alumni had roughly 16 weeks plus tenure. Cloudflare alumni have eight months of full base pay, so they ignore lowball offers and run parallel processes. Composition-wise, Cloudflare cut "measurers" (middle management, finance, legal) more heavily than "builders," so the engineering slice is smaller but more concentrated in agentic and platform functions than Oracle's broader infrastructure cuts.
What signals identify the agentic AI engineers inside the 1,100?
Look for engineers who worked on internal tooling, developer platforms, ops automation, or data plumbing at Cloudflare in 2025 and 2026, especially anyone whose GitHub activity shows orchestration frameworks, eval harnesses, or LLM gateway work. The cohort that got cut because their function was absorbed by agents is the same cohort that knows how to build and supervise those agents elsewhere. Headline keywords on LinkedIn are a weak signal; recent commit activity and internal team membership are stronger.
Should we wait for "Open to Work" badges before reaching out?
No. By the time this cohort flips the badge on (likely October or November) you're in a queue of dozens of recruiters. The right move is to reach out in August or September with a low-pressure intro that explicitly does not pitch a role yet. The goal is to be in their trusted inbox by the time they decide to start a process, not to convert them on first contact.