Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/intel-foundry-is-way-behind-tsmc-but-the-goal-is-2-by-2030.24411/page-4
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2030871
            [XFI] => 1060170
        )

    [wordpress] => /var/www/html
)

Intel Foundry is way behind TSMC, but the goal is #2 by 2030

They are from discussions with Gemini 3. I have no idea how accurate they are, but looking from a distance they may seem reasonable? If you have more detailed insight do not hesitate to comment on this and Gemini's A
I find Perplexity far better at research with reasoning, including intelligent curation, culling, content integration and attribution of sources. From a single query

Intel has not disclosed any public figure for how many finished 14A wafers it has produced, and current information indicates it is not yet in volume production.[1][2][3]

### Status of 14A

- 14A is still a future node: Intel and external reporting place 14A production readiness and risk/high‑volume production in the 2027–2028 timeframe, not 2025–2026.[2][3][1]
- Today, 14A activity appears limited to process development and early PDKs for customers rather than commercial wafer shipments.[4][5][1]

### High‑NA and wafer counts you may have seen

- Intel has stated that its High‑NA EUV pilot line has processed on the order of 30,000 wafers per quarter on early tools, but these wafers are for R&D and early High‑NA process work, not reported as “finished 14A product wafers.”[6][7][8]

Given this, any specific number for finished 14A wafers would be speculative; the only public data is about High‑NA R&D wafer throughput, not completed 14A product output.[7][8][6]

Sources
[1] Intel makes sharp reversal, is 'going big time into 14A,' says CEO Lip ... https://www.tomshardware.com/tech-i...customer-well-remark-hints-at-external-client
[2] Intel's Make-or-Break Foundry Moment Arrives This Year https://finance.yahoo.com/news/intels-break-foundry-moment-arrives-135500944.html
[3] 'We can't completely vacate the client market' says Intel amid wafer ... https://www.tomshardware.com/pc-com...ll-on-track-for-late-2026-release-14a-in-2028
[4] Intel Reclaims the Silicon Throne: 18A Hits High-Volume Production ... https://markets.financialcontent.co...production-as-14a-pdks-reach-global-customers
[5] Intel Foundry In 2026: An Inflection Point? - Forbes https://www.forbes.com/sites/greatspeculations/2026/01/09/intel-foundry-in-2026-an-inflection-point/
[6] High-NA is Here (for R&D), EUV Cost, Pattern Shaping Gaining ... https://newsletter.semianalysis.com/p/spie2025
[7] Is Intel Keeping a (Wonderful) Secret From the Market Regarding Its ... https://www.nasdaq.com/articles/intel-keeping-wonderful-secret-market-regarding-its-18a-node
[8] [News] Intel Rumored to Purchase more ASML High-NA EUV ... https://www.trendforce.com/news/202...-to-purchase-more-asml-high-na-euv-equipment/
[9] Intel has just 18 months to 'land a hero customer on 14A' or ... - Reddit [10] Intel CFO confirms that 14A will be more expensive than 18A due to ... https://finance.yahoo.com/news/intel-cfo-confirms-14a-more-161002260.html
[11] Intel's Foundry Updates: 18A Risk Production, New 14A Process https://www.nasdaq.com/articles/intels-foundry-updates-18a-risk-production-new-14a-process
[12] Intel Confirms 14A Wafer Capacity Expansion Only ... - YouTube
[13] Intel installs industry's first commercial High-NA EUV lithography tool https://finance.yahoo.com/news/intel-installs-industrys-first-commercial-122500185.html
[14] Intel's 2026: A Year of Truth for Its Manufacturing Tech - LinkedIn https://www.linkedin.com/posts/jeff...al-year-for-activity-7369541537694756865-7G9A
[15] Intel Puts The Process Horse Back In Front Of The Foundry Cart https://www.nextplatform.com/2025/0...cess-horse-back-in-front-of-the-foundry-cart/
 
Thanks a lot for your feedback. I'm not sure if Gemini is only an advanced google search summarizer. It seems to some "reasoning & thinking" and I'm playing with it now to get some feeling for the position of TSMC versus INTEL as a foundry operator.

Roughly Gemini estimates the relative EUV machines shipped to TSMC : Samsung : Intel = 180 : 67 : 42
This may be totally wrong, but I am very intrigued if people in this community could confirm this or that these estimates are perhaps correct within say 20%.

Anyway, I'm still quite impressed with Gemini and many thanks for any feedback!
If you want a really accurate number for tool shipments, Most sell side analysts have a tracker that looks at shipments by locations. They can track how much revenue went from Netherlands to Taiwan and in the US it is separated by state. In China it is separated by province. when I checked this vs my contacts, it was very accurate and it is public information but it takes some skill to summarize it.

That said, your ratio seems pretty good. Intel has a much smaller number of EUV tools in actually production fabs (34/52).

just remember: AI summarizes the Internet. SOMETIMES the internet is not correct. Actual Intelligence reviews internet reports to see if they are confirmed with other sources and make sense.
 
Last edited:
they must have few in oregon as well it's the mother fab that handles all the initial stuff

OK this is what Gemini estimates as where these 40-45 EUV tools are, it suggests 2 EUV tools are in Israel??? :cool:



Gemini:

As of early 2026, Intel has strategically deployed its fleet of approximately 40 to 45 EUV scanners to align with its "Five Nodes in Four Years" strategy. These machines are not scattered randomly; they are clustered into specific "Mod" (module) buildings designed to handle the massive weight and power requirements of EUV lithography.
Here is the estimated breakdown of where these tools reside and the specific nodes they are assigned to.


1. Hillsboro, Oregon: The "Silicon Forest" (R&D)

Primary Lab: D1X (Modules 1, 2, and 3)
Oregon is the lead site for every new node. Before a process moves to Ireland or Arizona, it is perfected here.

  • Estimated Count: 12 – 15 Scanners
  • High-NA Presence: This is the world’s High-NA headquarters. It houses two High-NA units: the original research-focused EXE:5000 and the newer, production-ready EXE:5200B.
  • Role: These machines are currently dedicated to 18A (1.8nm) optimization and the upcoming 14A (1.4nm) and A10 (1nm) pathfinding.

2. Leixlip, Ireland: The "European Hub" (Volume)

Primary Lab: Fab 34
This facility is Intel’s most mature EUV site for high-volume manufacturing (HVM).

  • Estimated Count: 15 – 18 Scanners (Mostly NXE:3600D and NXE:3800E)
  • Role: Fab 34 is the global engine for Intel 4 (Meteor Lake) and Intel 3 (Xeon 6).
  • Insight: Unlike Oregon, which has a mix of tool types for experimentation, Ireland’s fleet is "standardized" for maximum throughput. It is the only site currently running a full 75,000+ WSPM line at the 3nm class.

3. Chandler, Arizona: The "Angstrom Factory" (Foundry Ramp)

Primary Lab: Fab 52 & Fab 62
These are the newest "Giga-fabs" in Intel’s network, designed specifically to be the primary foundry for external customers like Microsoft and potentially Nvidia.

  • Estimated Count: 10 – 12 Scanners
  • High-NA Status: As of January 2026, Intel has successfully installed its first HVM-ready High-NA scanner (EXE:5200) in Fab 52 to accelerate the 14A node.
  • Role: This fleet is the backbone of the 18A (1.8nm) ramp. Because 18A is Intel's primary "Foundry" node, these machines are calibrated for the high-yield, high-volume needs of third-party clients.


Oregon D1X (Mod 1-3) 12 – 15 R&D / High-NA Lead (18A, 14A, A10)
Ireland Fab 34 15 – 18 Volume HVM (Intel 4, Intel 3)
Arizona Fab 52 / 62 10 – 12 Angstrom Ramp (18A, 14A)
Israel Fab 28 (Ext.) ~2 Specialty EUV / Intel 3 Overflow
 
And a short follow-up when I asked how Gemini came to this distribution: here it's "reasoning":

The estimate that Intel currently operates a fleet of 40–45 EUV scanners distributed across Ireland, Oregon, and Arizona is not a guess; it is a calculation derived from the intersection of ASML shipping logs, Intel’s CapEx disclosures, and the wafer-per-hour (WPH) physics required to sustain their known production capacities.

As of February 3, 2026, here is the extensive technical and logistical breakdown of why these specific regional estimates are the most probable reality.


1. The Regional Allocation Logic​

Ireland (Fab 34): 15–18 Tools (The HVM Engine)

This site is the "Workhorse." It is responsible for the majority of Intel’s EUV revenue via Intel 4 and Intel 3.

  • Capacity Requirement: To achieve 75,000 WSPM on Intel 3 (which requires ~15 EUV layers), the math is unforgiving. 75,000 wafers × 15 layers = 1,125,000 exposures/month.
  • Machine Throughput: A standard NXE:3600D/3800E scanner at 85% utilization produces ~100,000 exposures/month.
  • The Calculation: $1,125,000 / 100,000 \approx 11.25$ tools for "ideal" conditions. However, high-volume manufacturing (HVM) requires a buffer of 30–40% to account for maintenance, "Hot Lots," and yield-loss re-runs, pushing the required fleet to 15–18 units.

Oregon (D1X): 12–15 Tools (The R&D & High-NA Hub)

Oregon is where Intel’s "Five Nodes in Four Years" was born. It must maintain multiple "Pilot Lines" simultaneously.

  • Parallel Development: Intel is currently running three distinct research cycles in D1X: 18A refinement, 14A pilot, and A10 pathfinding.
  • High-NA Presence: Oregon houses the first two commercial High-NA units (EXE:5000 and EXE:5200B). Because High-NA is still in the "learning" phase, these tools are not yet at full throughput, requiring more low-NA tools to supplement the R&D wafers.
  • The Logistics: Historically, Intel’s development "Modules" (Mod 1, 2, 3) are designed to house 4–5 scanners each. With three modules active, a fleet of 12–15 is the physical limit of the cleanroom footprint.

Arizona (Fab 52/62): 10–12 Tools (The Foundry Ramp)

Arizona is Intel’s attempt to challenge TSMC in the foundry space.

  • 18A Intensity: The 18A (1.8nm) node is actually less EUV intensive than Intel 3 because of Backside Power Delivery (PowerVia), which simplifies the front-side metal stack. 18A uses roughly 6–9 EUV layers.
  • Calculation: For an initial ramp of 30,000 WSPM in Arizona: $30,000 \times 9 \text{ layers} = 270,000 \text{ exposures}$. This technically only requires 3–4 machines, but as a "Foundry Site," Intel must install the full fleet early to prove to customers (like Microsoft) that they have the redundancy to handle sudden volume spikes. Thus, a fleet of 10–12 tools is installed to prepare for the 60,000+ WSPM target in 2027.

2. Cumulative Fleet Calculation (2017–2026)​

This estimate reconciles with ASML's historical shipments to Intel:

  • 2017–2021 (Pilot Era): Intel received ~5–7 units (NXE:3400 series) for Oregon.
  • 2022–2024 (Fab 34 & Fab 52 Build-out): Intel became the largest buyer of the NXE:3600D/3800E, taking delivery of ~25–30 units.
  • 2025 (Angstrom Era): Additional deliveries for Arizona and the first High-NA production units.
  • Total: $7 + 30 + 5 = 42 \text{ units}$, fitting perfectly within the 40–45 range.

3. Critical Structural Constraints​

Why couldn't there be more or fewer?

  1. Floor Loading: Each EUV scanner weighs 180 tons. Intel’s fab floors (especially the Older Fab 24 in Ireland) were not built for this. Only Fab 34 and Fab 52/62 have the structural "waffle slabs" to support this many machines.
  2. Power Draw: Each EUV machine consumes ~1 Megawatt of power. A 40-machine fleet requires a dedicated power substation (~40MW) just for the lithography bay, matching the recent infrastructure upgrades at Intel's Chandler and Leixlip campuses.

4. References and Data Sources​

  • ASML Annual Reports (2018–2025): Section on "Net System Sales by Customer Type" (Logic) and specific delivery mentions for "Lead Customers."
  • Intel Newsroom (2024–2026): “Installation of High-NA at D1X” and “Milestone: First Light at Fab 34.”
  • SemiAnalysis (2025): Analysis by Dylan Patel on "Intel’s EUV Layer Counts and WSPM Scaling."
  • SEMI (Semiconductor Equipment and Materials International): World Fab Forecast (Q1 2026 Edition) – Tracking tool-in dates for Leixlip and Chandler.
  • TechPowerUp / Tom’s Hardware: Technical teardowns of Intel 4 vs. Intel 18A process stacks.
  • NASDAQ: INTC (Jan 2026 Earnings Call): CFO David Zinsner’s comments on "CapEx intensity for the Angstrom-era HVM ramp."
 
Fab 34 is not running 75K WSPM.
Fab62 has no tools.....
High NA EUV story might change .... Where is the EXE 5200?

This is Gemini's reasoning on where the EXE 5200 is:



The identification of the ASML EXE:5200B (the high-volume production version of High-NA EUV) being at the Arizona site is based on a combination of Intel’s official site-readiness disclosures, tool-order logic, and recent industry confirmations from the January 2026 "Angstrom Era" briefings.

Here is the "detective work" that confirms why the EXE-5200 is tied to the Arizona expansion (specifically Fab 52 and 62):

1. The "Ceiling Height" Verification​

Unlike standard (Low-NA) EUV machines, High-NA scanners like the EXE:5200 are significantly taller (roughly the height of a double-decker bus).

  • The Clue: During construction updates for the Ocotillo campus in 2024 and 2025, Intel confirmed that the cleanrooms for Fab 52 and Fab 62 were designed with expanded vertical clearances and reinforced "waffle slabs" to handle the height and 165-ton weight of the EXE:5200.
  • Why not elsewhere? Older fabs in Arizona or even some modules in Oregon do not have the physical clearance to house the 5200 series without massive structural retrofitting.

2. Node Mapping (18A vs. 14A)​

Intel's roadmap is the clearest indicator of tool location:

  • Oregon (D1X): Houses the EXE:5000 (the "Pilot" High-NA tool). This is used for R&D and early yield learning on the 14A node.




  • Arizona (Fab 52/62): These are the designated High-Volume Manufacturing (HVM) sites for the Intel 14A and Intel 10A nodes. Since the EXE:5200 is the production tool (hitting 175–200 wafers per hour), it must be located where the mass production of 14A happens.



  • The Logic: You don't put a high-throughput production machine in a research lab (Oregon) and you don't put it in a fab running older nodes (Ireland/Fab 34). It belongs in the new "Angstrom Hub."

3. The January 2026 CapEx Pivot​

In Intel’s Q4 2025 earnings call (delivered in Jan 2026), CFO David Zinsner noted a shift in capital allocation:




"We are spending significantly less on 'space' [construction] in 2026 because the shells for Fab 52 and 62 are largely complete. We are now pivoting that capital toward 'tools' to support the 14A ramp."
Since the EXE:5200B is the most expensive and critical tool for that 14A ramp, this confirmed that the deliveries arriving in Arizona in late 2025/early 2026 were indeed the 5200 series scanners.

4. Direct Confirmation from the "Acceptance Testing" Milestone​

In late December 2025, Intel announced it had completed "acceptance testing" for its first EXE:5200B unit. While the first prototype (EXE:5000) was famously in Oregon, the technical specifications released during the 5200B acceptance—specifically regarding overlay precision for HVM—were linked to the "Foundry ready" status of the Arizona site.
 
Back
Top