MDW Backstory: Update On Intel

It goes without saying that the price of a share of a stock is not related to the value of the company itself.

Ticker INTC today:

INTC, P/E: 4,577:

AI pops today, link here.

overall market sinks 800 points, 3.3% or worse.But Intel (INTC) beats the market by a huge margin.
why: Seaport analyst Jay Goldberg upgraded Intel to buy, highlighting the company's upcoming Panther Lake chips, which uses it's 18A manufacturing process.

****************************
2026 CES -- Last Week

Intel, of course, is very upbeat, very, very happy with progress on the 18A.

A reminder: when this all started, "everyone" agreed that the 18A was a stop-gap measure, while Intel continued work on the 14A -- Intel is betting the company on 14A

Concerns have been raised about Intel's success with their 18A process.

*************************
Background on Fab Production -- Process Yield

AI prompt: What is the accepted process yield considered acceptable to fabs?

AI reply:

In semiconductor fabrication, an "acceptable" process yield varies heavily depending on whether the technology is mature or in early development, but generally, above 90% is considered good for mature, high-volume production.

For cutting-edge or new process nodes, initial yields can start below 50% during risk production, with 70–80% required for stable, profitable mass production.

Key Yield Standards by Production Stage

Mature Technology (Volume Production): >90% is the standard expectation for profitability.

Stable Production/Wafer Test: >90% to >98% is expected, depending on die size and complexity.

New Node/Risk Production: <50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.="">
Factors Affecting "Acceptable" Yield

Die Size: Larger, more complex chips (e.g., high-end GPUs or server CPUs) have lower yields due to higher probability of hitting a defect, sometimes operating profitably with lower percentage yields compared to small, simple chips.

Defect Density

( D0 𝐷0 ): A mature, high-volume fab generally aims for a defect density of

D0<0 .1="" br="">
𝐷0<0 .1="" br="">
defects/cm².

Product Maturity: Yields are notoriously low at the start of a new technology node, representing a "learning" phase rather than a sustained production phase.

Industry Context While 100% yield is the ultimate goal, it is practically impossible due to microscopic defects and process limitations. A 90%+ yield is generally necessary for foundries to generate a return on investment (ROI) on expensive, advanced technology nodes.

<50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.=""><0 .1="" br=""><0 .1="" br="">Intel's 18A process yields have been a significant concern, with reports varying wildly, from low figures like 10-20% in late 2024/mid-2025 suggesting issues with mass production readiness, to more recent estimates (early 2026) placing them around 55-60%, indicating improving but still maturing levels, not yet at peak efficiency, but showing progress for future products like Panther Lake. Intel disputes the lowest numbers, acknowledging they aren't at full margin-driving yields but expect continuous improvement, with industry-standard yields targeted for 2027. Reported Yields Over Time (Varying Sources): Late 2024/Mid-2025: Reports suggested very low yields, around 5-10%, causing concern about mass production feasibility and disappointing potential clients like Broadcom. August 2025: Sources indicated yields improved to around 10%, with Intel disputing specifics but confirming ongoing improvement. Late 2025/Early 2026: Analyst reports placed yields closer to 55-60%, suggesting they're now good enough for initial product launches (like Panther Lake) but still behind leaders like TSMC. Intel's Stance: Intel officials acknowledge yields are improving monthly and are adequate for supply but not yet at optimal margin levels, with goals for industry-standard yields by 2027. Key Takeaway: The 18A process is maturing, moving from very low early yields to more viable levels, but remains a focus area for improvement to achieve cost-effective, high-volume production. Intel's 18A process yields have been a significant concern, with reports varying wildly, from low figures like 10-20% in late 2024/mid-2025 suggesting issues with mass production readiness, to more recent estimates (early 2026) placing them around 55-60%, indicating improving but still maturing levels, not yet at peak efficiency, but showing progress for future products like Panther Lake. Intel disputes the lowest numbers, acknowledging they aren't at full margin-driving yields but expect continuous improvement, with industry-standard yields targeted for 2027. Reported Yields Over Time (Varying Sources): Late 2024/Mid-2025: Reports suggested very low yields, around 5-10%, causing concern about mass production feasibility and disappointing potential clients like Broadcom. August 2025: Sources indicated yields improved to around 10%, with Intel disputing specifics but confirming ongoing improvement. Late 2025/Early 2026: Analyst reports placed yields closer to 55-60%, suggesting they're now good enough for initial product launches (like Panther Lake) but still behind leaders like TSMC. Intel's Stance: Intel officials acknowledge yields are improving monthly and are adequate for supply but not yet at optimal margin levels, with goals for industry-standard yields by 2027. Key Takeaway: The 18A process is maturing, moving from very low early yields to more viable levels, but remains a focus area for improvement to achieve cost-effective, high-volume production.

<50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.=""><0 .1="" br=""><0 .1="" br="">AI prompt: Intel says things are going very well; the company is happy with their 18A process. Putting everything together that you know, how is Intel and 18Areally going?

<50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.=""><0 .1="" br=""><0 .1="" br=""><50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.=""><0 .1="" br=""><0 .1="" br="">AI reply: Here’s a balanced assessment (as of Jan 20, 2026) of Intel’s claim — based largely on the strides with its 18A process(Intel’s 1.8 nm-class technology) — that it’s “well on its way to becoming a major fab competitor with TSMC”:

<50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.=""><0 .1="" br=""><0 .1="" br="">AI continues:

What Intel Actually Has Going For It:

1. Competitive 18A Process Performance (PPA gains). Intel’s 18A process uses advanced RibbonFET GAA transistors and backside power delivery, giving it potential performance and power-efficiency advantages versus its own previous nodes. Early technical comparisons suggest ~25 % performance improvement and ~30 % density gains over Intel 3 — and competitive edge metrics relative to TSMC’s N2 class on certain vectors.

2. Early volume production ahead of some rivals. Intel began high-volume manufacturing on 18A in 2025 — before TSMC’s N2 output reaches full grind — which Intel can tout as time-to-market leadership for certain products.

3. Strong financial & stock sentiment tailwinds. Wall Street optimism around Intel’s turnaround has grown, with analysts upgrading or pointing to improving fundamentals tied to the 18A launch and prospects for AI/compute growth — though a majority of analysts remain cautious.

4. U.S. fab leadership and strategic positioning. Intel’s Fab 52 in Arizona — making 18A parts — will be the most advanced fab on U.S. soil for several years, whereas TSMC’s comparable advanced production stateside lags until later in this decade.

Key Challenges and Reality Checks:

1. Execution & yield concerns still exist.

Reports suggest persistent yield and ramp issues and unresolved hurdles for high-end external customer adoption, meaning Intel’s ability to scale profitable foundry production remains unproven.

2. Foundry customers & ecosystem are thin.

Outside Intel’s own products and select internal clients (like Amazon, Microsoft, U.S. defense), there’s no broad roster of marquee external customers yet using 18A at volume — a core requirement to rival TSMC’s ecosystem.

3. TSMC’s ecosystem & volume advantage.

TSMC’s 2 nm-class (N2) process — with a deep existing customer base (Apple, AMD, Nvidia, Qualcomm) — continues to cement leadership in advanced logic, with mass production already underway or ramping.

TSMC also plans robust capex and fab expansions through the rest of the decade while maintaining share in AI/ HPC demand.

4. Strategic uncertainty around foundry roadmap.

At times there have been suggestions Intel might de-emphasize external 18A offerings and instead focus on the next 14A node — a move that could delay competitive posture.

Bottom Line — Where Intel Truly Stands:

Intel has narrowed the technology gap.

18A is competitive on certain technical benchmarks and delivered before some of TSMC’s equivalents. That’s meaningful progress after years of Intel lagging process leadership.

But it’s premature to call Intel a “major fab competitor to TSMC.”

TSMC still leads in:

Scale (volume & capacity),

Ecosystem breadth (major customers),

Risk, design support, and long-term node leadership.

Intel’s foundry business has momentum and strategic backing (even geopolitical), but matching TSMC as a global contract manufacturer by raw output or customer base remains a multi-year endeavor.

So, Intel isn’t an also-ran anymore — it’s a serious runner in the semiconductor fab race — but not yet a dethroner. The real inflection point will be yield maturity, cost competitiveness, and external customer adoption of 18A/14A over the next 2–4 years.

****************************
Intel 18A Update

<50 2nm="" 60="" 70="" a="" advanced="" br="" but="" commercial="" common="" considered="" e.g.="" for="" initially="" is="" must="" nm="" nodes="" rise="" strong="" target.="" toward="" viability.=""><0 .1="" br=""><0 .1="" br="">Intel's 18A process is a major advancement, debuting with the new Core Ultra Series 3 "Panther Lake" processors launched at CES 2026, bringing advanced features like RibbonFET (GAA) transistorsand PowerVia backside power delivery for improved performance and efficiency, enabling new AI PCs with powerful NPUs and strong battery life, with availability starting January 2026 for client devices and expanding for foundry customers, marking a significant step in Intel's manufacturing roadmap.

Key Highlights of 18A:

First to Market: It's the first sub-2nm node manufactured in the U.S. and integrates cutting-edge tech like Gate-All-Around (GAA) transistors (RibbonFET) and backside power delivery (PowerVia).

Performance Gains: Offers up to 15% better performance per watt and 30% greater density than Intel 3, with significant boosts in AI capabilities.

AI PC Focus: The new Core Ultra Series 3 chips built on 18A deliver enhanced NPUs for local AI processing, qualifying them for Co-Pilot+ PC experiences.

Broad Adoption: Powers over 200 PC designs and expands into edge/industrial applications, with availability starting January 2026.

Future Outlook:

Manufacturing Milestones: Intel aims to reach industry-standard yields for 18A by 2027, with future nodes like 14A and 14A-PT in development.

Foundry Growth: 18A is crucial for attracting external foundry customers, positioning Intel as a major player against TSMC.

Apple Rumors: There are ongoing rumors that Intel's 18A could eventually be used to manufacture some of Apple's chips, notes MacRumors.

MDW Backstory

Pages

Tuesday, January 20, 2026

Update On Intel - January 20, 2026

No comments:

Post a Comment

Report Abuse

Labels