fabintel

FAB INTELLIGENCE DOSSIER

Intel Ocotillo (Fab 52)

Chandler, AZ · 300mm Intel 18A logic · Generated 26 May 2026

Process / wafer

Intel 18A · RibbonFET + PowerVia · 300mm

Capacity at full ramp

~10,000 wafer starts / week (≈40k WSPM)

CHIPS Act exposure

$7.86B finalized direct funding (Nov 2024)

Signal strength

Active 18A ramp · public yield commentary

01Fab snapshot

7 fields
ParentIntel Corporation (NASDAQ: INTC)
SiteOcotillo campus, Chandler, AZ — Fab 52 is the fifth high-volume fab on the campus
ProcessIntel 18A (RibbonFET GAA + PowerVia backside power); also runs as proof site for 14A internal product
Products in rampPanther Lake (Core Ultra Series 3, client) and Clearwater Forest (Xeon 6+, server)
Designed capacity~10,000 18A wafer starts per week per Chandrasekaran, CNBC Nov 2025
CHIPS Act funding$7.86B finalized direct funding, Nov 26, 2024; plus $8.9B federal equity stake (Aug 2025)
StatusFab 52 declared fully operational Oct 9, 2025; 18A HVM declared Dec 2025; Fab 62 still to come ~2028

02Recent signals

Last 12 months

CFO Zinsner: 18A 'still early in its ramp' on Q1 FY26 call — yields better than expected but cost pressure is real

Apr 2026

On the April 23, 2026 Q1 FY26 call, CFO David Zinsner said: "Intel 18A is still early in its ramp, and rising input costs, especially in memory, present growing headwinds in the second half that we need to overcome." The same call disclosed that non-GAAP gross margin came in at 41%, ~650 bps ahead of guidance, with better yields on Intel 18A offsetting some of the higher costs incurred in early ramp . Translation for an account view: the ramp is working but every basis point of defect-density improvement at Fab 52 is showing up directly in margin guide.

Yield painEarnings call

Chandrasekaran (Intel Foundry CTOO/GM) tells CNBC: defect density improving month-over-month at Fab 52

Dec 2025

In the November 2025 CNBC tour of Fab 52 in Chandler, Naga Chandrasekaran said: "We are making yield improvements, defect density improvements, month-over-month and hitting our goals. So I believe we have turned the corner." He also confirmed Fab 52 is capable of more than 10,000 18A wafer starts per week, with more than a million square feet of cleanroom across five Arizona fabs connected by 30 miles of overhead track . CNBC also explicitly noted some 18A wafers have had defects, making for a lower number of usable chips per wafer — i.e. the public framing is yield-and-defects, not architecture risk.

Yield painPress release

Naga Chandrasekaran consolidates Intel Foundry under one operations leader (Feb 2026)

Feb 2026

After Kevin O'Buckley left for Qualcomm in February 2026, Intel Foundry is now headed by Naga Chandrasekaran, who as Chief Technology and Operations Officer now oversees both development of advanced process technologies and day-to-day execution of the global manufacturing network . This collapses the prior split where O'Buckley owned services and Chandrasekaran owned manufacturing/TD. For the Ocotillo account that means one decision-maker sits above the entire 18A ramp — and the same person who has been publicly committing to month-over-month defect-density gains.

New decision-makerPress release

Fab 52 declared fully operational and Panther Lake / Clearwater Forest both anchored to Ocotillo

Oct 2025

Both Panther Lake and Clearwater Forest, as well as multiple generations of products built on Intel 18A, are being manufactured at Fab 52 , and Fab 52 is Intel's fifth high-volume fab at its Ocotillo campus in Chandler, Arizona, part of the $100 billion Intel is investing to expand its domestic operations . Kevin O'Buckley (still at Intel at the time of the announcement) framed it on the record: "The new fab that we've built here is all about enabling capacity for that new technology... it's all about bringing these new technologies to market for the first time, exercising this new facility, this new team and all these new tools."

Capacity pressurePress release

$7.86B CHIPS award finalized; Ocotillo carries explicit milestone language

Nov 2024

Arizona is home to Intel's largest domestic manufacturing footprint, with four microchip plants in the state and two additional plants under construction in Chandler. The award supports Intel's investments in advanced chip fabrication and packaging at its Ocotillo campus, where two state-of-the-art fabs are currently being built, enabling production of the company's most advanced process technologies, including Intel 18A . Federal money was finalized Nov 26, 2024; the August 2025 federal equity stake added another $8.9 billion investment, primarily coming from grants promised under the CHIPS Act signed by President Joe Biden in 2022 . Both attach to 18A capacity milestones the Ocotillo team has to deliver.

FundingPress release

Active 'Defect & Yield Metrology Engineer' and OTF process-integration reqs posted out of Phoenix/Chandler

May 2026

Intel currently has Phoenix-AZ-based listings for a Defect & Yield Metrology Engineer at $134K–$255K and an Ocotillo Technology Fabrication Senior Process Integration Engineer at $136K–$193K , plus a separate OTF Yield Failure Analysis Engineer role that operates and manages advanced analysis tools such as TEM . The Defect & Yield Metrology JD calls out experience in micro-contamination management within semiconductor manufacturing and experience with semiconductor process flows and/or defect metrology assessment . These are the roles a SixSense buyer-side champion typically sits in.

Buying signalLinkedIn / careers page

03Engineering org map

Public profiles only
CP

Connor Phillips

Yield Defect Metrology Engineering Area Manager, Intel Corporation (Greater Phoenix Area)

ASU alum; LinkedIn lists tenure at Intel including diffusion / RTP process tool ownership before moving into the yield defect metrology org. Direct owner of the inline-defect-metrology surface that maps to a SixSense classifier deployment.

PC

Philip Campbell

Yield Engineering Manager, Intel Corporation

PhD Materials Science, Georgia Tech. Self-describes as managing engineers responsible for tool development for inline defect metrology and 'novel defect modes' across FEOL/BEOL/Packaging — scripts in Python/MATLAB/JMP for automated defect data analysis.

MR

Matt Russell

Staff Process Integration Engineer, Intel Corporation (Chandler, AZ)

25 years at Intel, Chandler-based since 2018. LinkedIn role description names ownership of full process flow for interconnects, 'investigation and mitigation of yield signals', and reviewing inline defect images by layer to drive process module fixes.

DA

Doreen Ahmad

Global Yield Process Integration / Device Development Engineer, Intel Corporation (Chandler, AZ)

Self-described scope is supporting Technology Development startups on the foundry-yield side under IDM 2.0, working with global yield teams on electrical-test parameter yield improvements and process-margin tightening for new tech roadmaps.

SS

Steven Shong

Sr. Director of Manufacturing, Intel Corporation (Chandler, AZ)

Senior manufacturing leader based at the Chandler site per LinkedIn; sits above the on-shift fab manufacturing org that owns day-to-day execution on the Ocotillo floor.

NC

Naga Chandrasekaran

EVP, Chief Technology & Operations Officer and GM, Intel Foundry

Joined Intel Aug 2024 from Micron (where he ran Technology Development). As of Feb 2026 also runs Foundry Services. Public face of 18A yield / defect-density commitments on CNBC Nov 2025 tour of Fab 52.

04Technical pain surface

Inferred · mapped to SixSense

Defect-mode classification on a brand-new GAA + backside-power stack with no copy-exactly precedent

Confidence: high

Intel 18A is the first node combining RibbonFET (GAA) and PowerVia (backside power) in HVM. Per Intel's own VLSI 2025 paper, 18A provides over 30% density scaling and a full node of performance improvement vs Intel 3 through an industry-first combination of advanced interconnects, GAA transistor architecture, backside power, and design co-optimization . New layer stack = new defect signatures the existing in-house classifier library was never trained on. CNBC explicitly noted in Nov 2025 that some 18A wafers have had defects, making for a lower number of usable chips per wafer , and Chandrasekaran is committing to monthly defect-density gains in public.

SIXSENSE ANGLE

No-code defect classifier trained per fab on the operator's own SEM/optical images — lets Connor Phillips' inline-defect-metrology team stand up classifiers for backside-power-specific defect modes in days instead of waiting for Copy Exactly-style cycles back to Hillsboro TD.

Every bp of yield is showing up in margin guide — and the CFO is calling it out by name

Confidence: high

Zinsner specifically called 18A out as a Q2 gross-margin drag on the Q1 FY26 call: "Our Q2 gross margin guide declines modestly from Q1 due to a meaningfully larger contribution from Intel 18A, still early in its ramp, and some inventory benefits in Q1 that aren't expected to repeat in Q2." Combined with better yields on Intel 18A offsetting some of the higher costs we always incur in the early part of ramping a new node , this is a fab where defect-density velocity is now a board-level metric, not a fab-level metric. The org will fund anything that shortens classifier authoring time.

SIXSENSE ANGLE

Closed-loop yield feedback tying defect classifications back to specific process / tool / chamber / lot signals — collapses the cycle between an excursion at Ocotillo and the corresponding margin commentary the CFO has to make on the next call.

CHIPS milestones + DoD Secure Enclave compress the timeline for inspection-logic tuning

Confidence: medium

The $7.86B CHIPS award is tied to capacity and process milestones at Ocotillo ( enabling production of Intel 18A ), and Fab 52 sits inside the Secure Enclave program — by late 2025 reporting, a $3 billion DoD initiative executed through Intel's Arizona facilities, with the Ocotillo campus described as the only US site producing sub-2nm defense-grade chips for F-35 and advanced radar programs . Defense-grade chips don't tolerate downstream defect escapes, which raises the stakes on inspection-rule tuning during the ramp window.

SIXSENSE ANGLE

Configurable inspection logic — engineers at Ocotillo can author and tune ROI / overlay / classification rules per layer without waiting on a vendor retraining cycle, which matters most exactly during the period the CHIPS and Secure Enclave milestones are being scored.

Active reqs name 'defect metrology assessment' and yield FA on TEM as open positions out of Phoenix

Confidence: medium

Intel's currently-posted Phoenix-AZ Defect & Yield Metrology Engineer JD calls out experience with semiconductor process flows and/or defect metrology assessment based on process requirements , and a separate OTF Yield Failure Analysis Engineer role operates and manages advanced analysis tools such as TEM . The team is hiring to solve defect classification and FA in-house — i.e. they have an internal program, an internal owner, and unfilled headcount. That's the textbook moment to put a no-code classifier in front of the existing team rather than wait for a fresh hire to come up to speed.

SIXSENSE ANGLE

No-code classifier authored by the fab's own engineers — the same JD requirements ('process flows', 'defect metrology assessment') match the user persona SixSense's product is designed to make productive on day one rather than month six.

05Publications & talks

Last 18 months

Intel 18A Platform Technology Featuring RibbonFET (GAA) and PowerVia for Advanced High-Performance Computing

Intel Corporation (incl. K. K. Bhuwalka, A. A. Farid, J. S. Ayers et al.) · 2025 IEEE Symposium on VLSI Technology and Circuits, Kyoto, Japan (Paper T1-1)

06Suggested first-touch angle

Open with Connor Phillips, not Chandrasekaran. Chandrasekaran is the public quote machine for 18A yield, but the person who actually owns the inline-defect-metrology surface where a SixSense classifier lives is the Yield Defect Metrology Engineering Area Manager out of Greater Phoenix. He sits one layer above the engineers Intel is currently trying to hire under the Phoenix 'Defect & Yield Metrology Engineer' req, which means he is both the user-buyer and the person feeling the pain of unfilled headcount on a node that has month-over-month defect-density commitments attached to it.

Lead with a node-specific technical question, not 'how is the ramp going'. The RibbonFET-plus-PowerVia stack disclosed in Intel's VLSI 2025 paper introduces backside-power defect modes that the prior FinFET classifier libraries were never trained against — ask Connor whether his team is standing up the new classifiers on the inline SEM/optical side, or on the post-CMP defect inspection side after backside reveal. That question only makes sense from a peer.

Timing argument: the Q1 FY26 call (April 23, 2026) named 18A explicitly as a Q2 GM drag, and Chandrasekaran's CNBC quote ('month-over-month defect-density improvements') is now a commitment, not an aspiration. Plus, with O'Buckley out and Chandrasekaran consolidating both TD and manufacturing under one role (Feb 2026), the decision authority above Connor just collapsed from two leaders to one — so any tool decision that used to need cross-org alignment now needs one less sign-off. The next two quarters are when defect-classifier velocity gets graded against the CFO's gross-margin guide.

DRAFT EMAIL TO CONNOR PHILLIPS

Subject18A backside-power defect modes — classifier authoring at the OTF inline metrology layer

Connor — saw the Chandrasekaran CNBC quote out of the November Fab 52 tour about 'month-over-month' defect-density improvements on 18A, and then the Q1 call where Zinsner called 18A out specifically as the meaningful Q2 gross-margin contributor still early in its ramp. The pressure on your layer of the org is pretty visible from outside.

The question I keep coming back to is the new defect modes the RibbonFET + PowerVia stack from the VLSI 2025 T1-1 paper introduces — specifically the post-backside-reveal layers. When your team adds those classifiers, are you running them on the inline SEM / optical side, or on the post-CMP inspection side after backside thinning? The two paths have very different authoring-velocity profiles and I'd love your read on which one is actually the bottleneck at Ocotillo today.

Quick context on why I'm asking: we work with fab yield-engineering teams on the layer where the engineers themselves author and tune the classification logic — no vendor retraining cycle in the middle. The product is most useful exactly in the window you're in now, where a new layer stack is generating defect signatures the prior library doesn't cover.

Happy to share a short writeup on how a GAA-stack fab handled the post-backside-reveal classifier ramp without re-spinning the inspection recipes, if it's useful. Either way, would value your read on where the actual bottleneck is.

[Your name]