Cisco's May 14, 2026 WARN filings do something most restructuring notices do not: they name the exact titles Cisco no longer wants. 56 software engineers, 39 SWE tech leads, 12 directors of software engineering, 7 SREs, 7 QA engineers, plus 17 engineering product managers and 15 software engineering leaders across San Jose, Milpitas and San Francisco. In the same quarter Cisco reported $1.9B of hyperscaler AI infrastructure orders, more than triple the $600M booked a year earlier, and framed the cuts as a realignment around silicon, optics, security and AI, not savings.
If you got a notice, the read is not "Cisco is shrinking." The read is: Cisco is trading the stack you shipped for the stack it is now selling, and your resume still describes the old one.
What the WARN filing actually says about your next job
The eliminated roles sit inside classical routing, switching and IOS-XR software; the growing headcount sits inside Silicon One custom chips, Acacia coherent optics, and GPU-cluster fabric. That is the lane the Cisco WARN 2026 filings tell you is contracting versus expanding at every vendor like it.
The California filings, delivered May 14 with terminations starting July 13, 2026, eliminate 236 positions in San José, 154 in Milpitas, and 81 in San Francisco. The role-level breakdown:
- 56 software engineers
- 39 SWE tech leads
- 17 engineering product managers
- 15 software engineering leaders
- 12 directors of software engineering
- 12 engineering program managers
- 9 directors of product management
- 8 business ops managers
- 7 SREs, 7 program managers, 7 software QA engineers
Cisco announced fewer than 4,000 roles cut globally, less than 5% of its ~80,000 workforce, at Q3 FY26 earnings on May 13, 2026. The 10-K describes the Fiscal 2026 Plan as investment in silicon, optics, security and AI, with pre-tax charges of up to $1 billion. That is the frame Wall Street read. Your resume should read the same way.
Up from $600M a year earlier in a single quarter; Cisco expects AI infrastructure revenue to grow ~88% to $7.5B in FY27.
The three pools, and why the smallest one pays best
There are three pools a displaced Cisco engineer can land in, and they are not the same size. In Refolk's index, silicon and optics is roughly 3x scarcer than legacy networking SWE work, which is exactly why it prices higher.
| Segment | US pool | Source |
|---|---|---|
| Software engineers listing networking/routing/switching skills | 4,610 | Refolk index |
| Silicon/optics/ASIC/SerDes/photonics engineers (all titles) | 1,466 | Refolk index |
| Engineers with GPU-fabric skills (CUDA/RDMA/InfiniBand/NCCL) | 17,436 | Refolk index |
| AI-fabric pool ÷ networking-SWE pool | 3.8x | Derived |
| Networking-SWEs ÷ silicon/optics specialists | 3.1x | Derived |
| Meta share of sampled silicon/optics specialists | ~48% | Derived (12 of 25) |
Three things fall out of this table:
- The generic "AI engineer" pivot is the crowded one. 17,436 US engineers already list CUDA, RDMA, InfiniBand or NCCL. If you rewrite yourself into that pool without a differentiator, you are competing with model-serving people, ML infra people, and ex-quant HPC people who have been living there for years.
- The silicon/optics lane is the scarcest. 1,466 US specialists. If you have ever touched ASIC bring-up, NIC drivers, SerDes tuning, or optical DSP firmware, the resume that leads with those five words beats the resume that leads with "distributed systems."
- Hyperscalers own the demand side. In the sampled silicon/optics set, ~48% currently work at Meta. That is not a coincidence; that is where the buildout is.
The mechanism is boring supply and demand. Cisco took $9.3B in FY26 orders (with $4B in Q4 alone), roughly 60% Silicon One systems and 40% optics. Those orders typically lock customers in for three or more years. Every vendor selling into the same buildout is bidding for the same 1,466 specialists.
Rewriting a Cisco networking resume into a silicon/optics resume
Rewrite the bullets in the vocabulary of the buyer, not the seller. The buyer is a hyperscaler infrastructure org, and the words on their JDs are Silicon One, P200, 8223, 51.2 Tbps, Acacia, coherent DSP, scale-across, RDMA, NCCL, InfiniBand, GPU cluster interconnect.
The words to drop
- "IOS-XR feature development"
- "carrier-grade routing"
- "MPLS/VPN control plane"
- "team of 40 across 3 geos"
- "10 years of Cisco networking experience"
None of these are wrong. They are just describing the shrinking pool. A recruiter scanning 200 resumes for a Meta network ASIC role does not stop at any of them.
The words to add (only if true)
- Silicon One P200 / Cisco 8223 (51.2 Tbps): the specific product family displaced Cisco engineers should reference by name if they touched it. Cisco expected to ship its one-millionth Silicon One chip in Q2 FY26. Anyone who worked on P100/P200 bring-up, verification, or driver development has a resume worth two of the generic ones.
- Acacia coherent optics: a "worked on Acacia coherent DSP" or "400ZR/800ZR pluggable qualification" line item is worth more on a hyperscaler resume than "10 years IOS-XR."
- Scale-across fabric: the exact term Cisco uses for inter-datacenter GPU traffic. CEO Chuck Robbins said agentic AI queries generate "up to 25 times more network traffic than chatbots." That is the demand thesis; your bullets should reference the traffic pattern.
- RDMA, RoCEv2, NCCL, InfiniBand: the four keywords that separate a network engineer from a GPU-cluster network engineer. Even one production incident debugging NCCL over RoCEv2 is a bullet.
- Cost-per-bit, power-per-port, BOM reduction: Cisco's gross margin fell 260 bps to 66.0% as AI hardware became a larger part of the mix. Hyperscalers negotiate hard. Engineers who can put a number on cost or power savings outbid engineers who show only feature velocity.
This is the exact rewrite Refolk does automatically: paste the Meta or Arista posting, and Refolk pulls the matching bullets out of your Cisco history (ASIC bring-up you did in 2019, the Acacia integration you led in 2022) and drops the MPLS bullets that no longer help. It also scores how well you actually fit, which matters when the posting looks close but the JD hides a hardware requirement you cannot meet.
The 12 directors: stop describing spans, start describing tape-outs
If you were one of the 12 directors of software engineering cut, the resume mistake to avoid is leading with headcount. In a quarter where Cisco product orders climbed 35% YoY and even ex-hyperscaler orders rose 19%, directors were not cut for demand reasons; they were cut because their orgs owned the wrong stack.
The rewrite for a director-level Cisco layoff resume is to lead with shipped artifacts, not org charts:
- Replace "led team of 42 across San Jose and Bangalore" with "shipped v2 of [product] on [ASIC family], qualified at [customer], first-year revenue $Xm."
- Replace "responsible for IOS-XR feature roadmap" with "owned silicon-to-software integration for [chip], including driver, firmware, and NOS layers."
- Replace "managed 6 engineering managers" with "ran three tape-outs and two optics qualification cycles in 24 months, hit A0 silicon on two."
Hyperscalers hire directors who have shipped hardware-adjacent product. Legacy vendors hire directors who have managed people who ship software. The two resumes look nothing alike.
The 12 directors were not cut for demand. They were cut because their orgs owned the stack Cisco is retiring.
Where to actually send it: Arista, Nvidia, and the neoclouds
The obvious targets are the two firms that took hyperscaler share from Cisco during the AI boom. The less obvious ones, and the ones with less resume competition, are the neoclouds.
- Arista Networks and Nvidia: the two named beneficiaries of Cisco's lost hyperscaler share. Both have been absorbing exactly this profile.
- Meta, Google, AWS, Microsoft: the four hyperscalers Cisco named in triple-digit AI order growth. They are also the biggest direct-hire competition. Meta in particular has been building a custom silicon and optics org; ~48% of sampled US silicon/optics specialists in Refolk's index already work there.
- Lambda, CoreWeave, and other "neoclouds": Cisco called out a separate $2B AI connectivity pipeline with neocloud, sovereign, and enterprise customers. These employers are smaller, less crowded with applicants, and hire the same profile. If you want fewer competing resumes, this is the lane.
Juniper (being absorbed by HPE) and HPE itself are not net hirers of this skill set. Sending a hyperscaler optics engineer resume to a shrinking legacy vendor is the same mistake Cisco just made at the org level.
Two audiences read your resume. Write for both
The recruiter and the ATS see different things, and your Cisco WARN 2026 rewrite has to satisfy both. The recruiter needs to see hyperscaler-vocabulary bullets in the top third; the ATS needs the exact keywords from the posting.
For every application, three things have to line up:
- Title mapping: if the JD says "Network ASIC Engineer" and your last title was "Principal Software Engineer, Routing," rewrite the title line to "Principal Engineer, Routing Silicon / NOS." Do not invent a role you did not hold; do rename the one you did hold in the vocabulary of the buyer.
- Skill list precision: if the JD names RoCEv2, put RoCEv2 in your skills, not "RDMA over Ethernet." ATS matching is literal.
- Story bullets: three to five bullets that tell a shipped-hardware story with a number attached. "Reduced port power 14%," "cut BOM $3.20 per line card," "achieved first-pass silicon on 5nm ASIC."
Doing this by hand for 40 postings is where most laid-off engineers give up and go back to Easy Apply. Refolk writes the tailored version of each application from your own history, drafts the cover letter to the same posting, and scores fit before you send.
The scarcer lane pays more and has fewer applicants; if you touched ASIC bring-up or optical DSP, lead with it.
The 3-year floor, not a Q2 window
Write your resume for a three-year hiring cycle, because that is how long the buyer contracts are. Hyperscaler AI orders typically lock customers in for three or more years, and Cisco alone expects AI infrastructure revenue to grow ~88% to $7.5B in FY27. Every vendor selling into that buildout is going to be hiring silicon, optics, and GPU-fabric people for the same window.
That changes two things:
- Do not chase a hot posting. Chase the skill adjacencies that will be hot for 36 months. ASIC verification, coherent optics, RoCEv2, and NCCL tuning are all in that bucket.
- Do not accept a role that puts you back in the shrinking lane. A "senior networking SWE" title at another legacy vendor in 2026 is a two-year reprieve, not a career. The same restructuring pattern is coming to every vendor that has not yet made the pivot.
Cisco layered the 2026 cuts on top of over 13,000 positions eliminated between early 2024 and end of 2025. The pool of displaced Cisco alumni is already deep. What is not deep is the pool of Cisco alumni whose resumes read like hyperscaler resumes. Be one of those.
FAQ
I did 12 years of IOS-XR and never touched silicon. Am I locked out?
No, but the pivot is longer. The realistic path is through GPU-fabric adjacencies (RoCEv2, RDMA, NCCL) rather than a cold jump into ASIC design. Look for postings titled "Datacenter Network Engineer, AI Infrastructure" or "Network Software Engineer, GPU Cluster" at Meta, Nvidia, CoreWeave and Lambda. Your control-plane experience translates; your MPLS/VPN experience does not. Rewrite the bullets around traffic patterns, congestion control, and lossless Ethernet, and drop the SP routing history to a two-line "prior experience."
How do I signal Silicon One or Acacia work if it was only one project?
Name the product and the ship date. "Contributed to Acacia 400ZR pluggable qualification, shipped Q2 2023" is worth more than a paragraph of generic optics prose. Recruiters at Arista, Nvidia and the neoclouds search on product names. If you spent six months integrating Silicon One into a customer NOS, put "Silicon One" in your title line or your top three bullets, not buried in paragraph four.
Should I mention the Cisco layoff on my resume or LinkedIn?
Not on the resume. The end date does the work. On LinkedIn, a one-line "Open to Silicon One / Acacia / hyperscaler AI infrastructure roles after Cisco's Fiscal 2026 restructuring" headline is fine and actually helps sourcers filter for you. What does not work is a long post about the layoff itself; recruiters at the target companies already know, and the algorithm rewards the specific-role signal, not the sentiment.
The JDs at Meta and Nvidia want 5+ years of ASIC experience I do not have. Now what?
Apply anyway to the ones where you meet 60% of the requirements, and target the neocloud and sovereign-cloud postings harder. Cisco's own $2B pipeline with neoclouds and sovereigns tells you those buyers are hiring the same profile with less credential inflation.