brain/
questionopenstock-market

Does Samsung have a credible leading-edge foundry comeback that would weaken the Intel-as-alternative thesis?

Notes

Does Samsung have a credible leading-edge foundry comeback that would weaken the Intel-as-alternative thesis?

The question

The us-fab-capacity-bottleneck thesis frames the world as bilateral: TSMC (capped) vs Intel (re-rating). But TrendForce reporting noted Apple is "eyeing Samsung, Intel U.S. foundry for core chips." If Samsung's leading-edge foundry execution catches up — yields, US footprint, customer wins — the Intel anchor-customer narrative dilutes (Apple, AWS, MSFT can split the spoils three ways instead of bestowing them on Intel alone).

Why it matters

A confident "Intel alone is the alternative" thesis supports a sharper INTC re-rate. A "Samsung is also legitimate" world means Apple / hyperscaler defection from TSMC happens, but spreads across three foundries — INTC's share of that flow is materially smaller. The thesis still works directionally; it just resizes.

What we currently believe

  • Q2 2026 did not make foundry a disclosed profit center. DS printed KRW 127.5T revenue / KRW 89.2T operating profit; memory is the engine (trade recaps: 99.7% of group OP). Official foundry language is "improved significantly prior to incentive-related provisions," with HBM base-die demand and U.S. orders as the drivers. Turnaround is "possible in the near term," not dated. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • 2nm yield contradiction (do not silently retire April ~55%). January "70% SF2P" is still on the wiki as a test-lot claim. April TrendForce ~55% remains the May 27 source of record. August Chosun/TrendForce now estimate ~60%, still below the ~70% bar Qualcomm/AMD are said to need. Official Q2 call gave no yield percentage. Recorded, not merged. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • Utilization recovered on nodes Samsung can already run: official 8nm-and-below at full; Chosun blended foundry 70–80% aiming at 100% in 2H. Advanced-node revenue guided >50%; AI/HPC mix high-teens → >30%. Price hikes of 10–15% on SF4/SF5/8nm are utilization/scarcity, not a 2nm yield victory. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • Qualcomm 2nm flagship still TSMC. Snapdragon 8 Elite Gen 6 / Gen 6 Pro reported exclusive to TSMC N2P because Samsung has not cleared a 70% quality bar. Qualcomm is already an SF4 customer. The May "still mulling" line did not become a 2nm flagship win. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • AMD Venice is official TSMC 2nm (Taiwan ramp; Arizona later). The May "AMD evaluating SF2P for Venice" line is not a closed Samsung order in any fetched post-May primary. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • The new named 2nm customer is a July 25 Broadcom MOU the companies "expect" at more than $200B across memory and foundry through 2030 (2nm and below + packaging). Framework, not a disclosed wafer contract. Q2 call: Broadcom talks ongoing; unnamed CSP / AI-HPC 2nm projects in design; 2nm project wins expected to more than double YoY. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • Apple leading-edge SoC still open. This pass fetched no public Apple–Samsung acknowledgement or repudiation after May 27. Exploratory Taylor visit (May 5) remains the last cited fact. Do not treat 2025 Austin CIS/image-sensor stories as a 2026 A/M-series award.
  • High-NA is not a 2027 volume allocation. Two tools installed; commercial High-NA delayed to 1nm-class (~2030) per NGL 2026. Intel is already using High-NA on selected 18A layers. Taylor Fab 1 operations 2026 / gradual 2nm ramp; Fab 2 groundbreaking late 2026, mass production 2030. From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative.
  • Thesis read after Q2: Samsung is a more credible 4nm / HBM-base-die / utilization third, and has a named 2nm HPC framework customer. It is not yet a co-equal leading-edge alternative that would split Apple / Qualcomm flagship / AMD Venice volume. The Intel-as-alternative thesis is resized at the margin, not falsified. Question stays open.

Evidence we have

  • 2026-05-08-autoresearch-intel-foundry-anchor-customers — TrendForce mention of Apple evaluating Samsung; 2027 ASML allocation (Samsung 7 Low-NA, Intel High-NA leadership).
  • From 2026-05-11-autoresearch-samsung-foundry-third-alternative-2026: Samsung SF2P yield 70% (January 2026); 4 years of GAA transistor telemetry; Taylor fab 90% ready (April 16, 2026); Q2 2026 initial 2nm manufacturing start; 50K WPM target, 100K by 2027; second Taylor fab under review. Qualcomm/AMD in "final negotiations" for Samsung 2nm.
  • From 2026-05-11-autoresearch-samsung-foundry-third-alternative-2026: Apple executives physically visited Taylor facility (Bloomberg, May 5, 2026) — the same week Apple-Intel preliminary deal was reported.
  • From 2026-05-11-autoresearch-intel-18a-yield-vs-tsmc-samsung-2026: Three-way comparison: Intel 18A ~55–65%, Samsung SF2P ~70%, TSMC N2 ~65–75%. Samsung at 70% is currently within striking distance of TSMC.
  • From 2026-05-11-autoresearch-picks-shovels-semicap-update-may-2026: Samsung 2 High-NA EUV units in 2026 — equal to Intel's allocation. Samsung is investing at leading-edge tool cadence, not an also-ran.
  • From 2026-05-18-autoresearch-samsung-foundry-as-third-alternative-may-2026: SF2P at 70%; 4nm >80% yield; utilization highest-ever; profit path Q2 2026. Digitimes May 12: Intel's Apple deal "puts Samsung foundry ambitions under pressure" — Apple engagement remains exploratory. Qualcomm CEO Korea visit April 2026. Tesla AI6 $17B Samsung deal confirmed (automotive inference, separate from Tesla Austin Intel 14A R&D fab). OpenAI HBM4 supply deal. SF2P+ announced; 1.4nm pushed to 2028–2029.
  • From 2026-05-20-autoresearch-samsung-foundry-q2-2026-profit-2nm-qualcomm-risk: SF2 vs SF2P yield divergence: SF2P (second-gen, Apple-grade) at 70%; SF2 (first-gen, mass-market) at ~55% per TrendForce April 14 — below mass-production threshold. Qualcomm defection risk: TrendForce Apr 14 flags Qualcomm "may opt for TSMC" if SF2 yields don't improve, losing a major 2nm anchor customer and making Samsung's 2T won FY2026 profit target harder.
  • From 2026-05-20-autoresearch-samsung-foundry-q2-2026-profit-2nm-qualcomm-risk: Samsung Q1 2026 DS profit +48x YoY — but this is HBM-driven (AI memory boom), not foundry. Foundry's standalone profitability is not publicly disclosed; 2T won is an internal goal. Foundry targets full 2nm utilization Q2 2026 and H2 2026 mass production of SF2P for mobile.
  • From 2026-05-22-autoresearch-samsung-foundry-may-22-2026: Qualcomm still mulling, not decided: April 23, 2026 Digitimes — "Qualcomm mulls return to Samsung 2nm as TSMC targets its LPU market." WCCFTech: full flagship switch "impossible" on timing; dual-source (lower/mid-tier Snapdragon to Samsung, flagship to TSMC) most likely path. Qualcomm CEO Amon confirmed Samsung talks at CES January 2026. AMD nearing a deal (TrendForce Dec 2025; unconfirmed as of May 22).
  • From 2026-05-22-autoresearch-samsung-foundry-may-22-2026: Samsung Q1 2026 consolidated record: KRW 133.9T revenue, KRW 57.2T op profit (both all-time records). DS Division KRW 81.7T revenue / KRW 53.7T op profit — entirely HBM-driven. Foundry earnings declined seasonally in Q1; full utilization target Q2 2026; H2 2026 SF2P mass production ramp.
  • From 2026-05-22-autoresearch-samsung-foundry-may-22-2026: TSMC/LPU competitive angle: TSMC is increasingly targeting the LPU market that overlaps with some of Qualcomm's addressable market — this gives Samsung structural leverage in Qualcomm negotiations (Samsung as TSMC-alternative for mid-tier Snapdragon to protect Qualcomm from TSMC's competitive expansion).
  • From 2026-05-27-autoresearch-samsung-foundry-third-alternative-may-2026: Yield clarification (April 2026): TrendForce April 14 reports Samsung 2nm yields at ~55% overall — below the ~70% mass production threshold. The January "70% yield" claim for SF2P appears to have been from a specific test-lot result, not sustained production average. The more recent (April) TrendForce data is authoritative. Qualcomm "may opt for TSMC" at current yield levels.
  • From 2026-05-27-autoresearch-samsung-foundry-third-alternative-may-2026: Samsung Chairman Lee Jae-yong visits MediaTek (May 21, 2026): Flew to Taiwan for direct talks at MediaTek HQ; pitch is bundled memory + foundry — Samsung HBM + Samsung Foundry manufacturing for MediaTek's next SoC, offered as integrated stack. MediaTek is TSMC's largest foundry customer by volume. This is the most aggressive customer-poaching move yet; 55% 2nm yield makes MediaTek conversion unlikely in 2026 but signals Samsung's strategic intent.
  • From 2026-05-27-autoresearch-samsung-foundry-third-alternative-may-2026: AMD evaluating Samsung 2nm (UPI May 6, 2026): AMD California reportedly evaluating SF2P for "Venice" EPYC server CPUs. If AMD moves even partial volume to Samsung, it would be the most significant non-Apple 2nm foundry win and evidence that SF2P can support complex CPU die.
  • From 2026-05-27-autoresearch-samsung-foundry-third-alternative-may-2026: Q1 2026 utilization 80% — one-year high, driven by mature-node (4nm, 5nm) client wins, not 2nm. Samsung 4nm yield at 80% (mature). SF1.4 confirmed pushed to 2029 (from prior 2028 guidance). Samsung optical module foundry win (May 1); mass production H2 2026.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: Q2 2026 official (newsroom + call): consolidated KRW 171.5T / OP KRW 89.5T; DS KRW 127.5T / OP KRW 89.2T. Foundry improved pre-provision on HBM base dies + U.S. orders; 2nm HPC design wins; H2 SF2P mobile + 4nm LPU/base-die; advanced-node revenue >50%; AI/HPC mix >30%. Official: 8nm-and-below full utilization; turnaround "near term"; no yield %. Taylor Fab 1 2026 / Fab 2 groundbreaking end-2026, MP 2030; 1.4nm inquiries, extra space under review.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: Yield series after May 27. TrendForce 2026-07-17 (Chosun Ilbo) still ~55%; TrendForce 2026-08-03 and 2026-08-21 (Chosun Biz / Seoul Economy) ~60%, bar still ~70% for Qualcomm/AMD scale. Contradiction vs May 27 ~55% and vs January 70% — recorded on this page; not silently closed.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: Qualcomm — WCCFTech/Business Korea: Gen 6 / Gen 6 Pro exclusive TSMC N2P; Samsung 2nm stuck ~60% vs 70% bar. TrendForce 2026-08-21: Qualcomm already an SF4 customer; Google in SF4 talks; Groq-3 LPU lifting 4nm.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: AMD — official 2026-05-21 Venice ramp is TSMC 2nm Taiwan + future TSMC Arizona. No fetched post-May Samsung close.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: Broadcom MOU (official 2026-07-25) — companies "expect" >$200B memory+foundry through 2030; 2nm and below + 2.3D/2.5D packaging. MOU / "estimated," not a booked wafer contract.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: High-NA — two tools installed; volume production delayed to 1nm-class ~2030 (NGL 2026 / Park Chang-min). Intel already using High-NA on selected 18A layers. 2027 High-NA scale test: no.
  • From 2026-08-24-autoresearch-samsung-foundry-q2-as-a-third-alternative: Apple leading-edge SoC — no fetched acknowledgement or repudiation after May 27.

Evidence we need

  • Public Apple-Samsung leading-edge SoC acknowledgement or repudiation (still missing after Q2)
  • Samsung Foundry standalone customer revenue / operating profit (still unpublished; Q2 is pre-provision + DS-consolidated)
  • Whether Qualcomm 2nm dual-source or AMD non-Venice volume actually closes
  • Binding Broadcom offtake (node mix, HBM vs foundry split, tape-out dates) behind the $200B MOU
  • ASML 2027 High-NA customer-split table — Samsung volume adoption now looks like ~2030, but allocation vs Intel/TSMC is still unfetched

How to resolve

  • Watch Samsung's quarterly earnings for a foundry-segment P&L, not just DS consolidated
  • Watch TrendForce / Korean press for whether 2nm yield prints ≥70% (the bar this page uses) rather than another 55–60% estimate
  • Watch for Apple supply chain signals (teardowns, supply-chain analyst coverage) indicating Taylor, TX as a production origin
  • Watch Broadcom project tape-outs / HBM qualification against the July 25 MOU

Related

Referenced by
Concepts
TSMC capacity shortfall and 4-year pricing powerUS fab capacity bottleneck → Intel as the alternative foundry
Questions
Has Intel 18A reached yield parity with TSMC, or are partnerships outpacing the technical reality?Will the Apple-Intel deal convert to actual orders, on which products, on what timeline?
Entities
Advanced Micro Devices (AMD)BroadcomIntelSamsung FoundryTSMC (Taiwan Semiconductor Manufacturing Co.)
Sources
Autoresearch: Samsung foundry Q2 as a third alternative to Intel/TSMC