Intel Foundry: the 18A node gamble

The facts

Intel is undergoing the most radical transformation in its history. Since March 2025, new CEO Lip-Bu Tan — a semiconductor veteran with decades of experience as an investor and chairman of Cadence Design Systems — has taken the helm of a company in deep crisis.

The context is brutal. Intel, which for decades was the undisputed leader in chip manufacturing, has lost ground to TSMC and Samsung in advanced process nodes. Chronic delays in transitioning to 10nm and 7nm nodes allowed AMD to erode significant market share. Tan's predecessor, Pat Gelsinger, had launched the ambitious IDM 2.0 plan to transform Intel into a world-class foundry, but was forced to resign in December 2024 after disappointing results.

Tan is betting everything on the 18A process node (equivalent to 1.8nm), which represents Intel's attempt to return to technological parity with TSMC. The reorganization has been drastic: 12 management layers compressed to 5, a corporate culture reoriented around the "bad news first" principle — bad news must travel up the chain of command quickly instead of being hidden.

The Q1 2026 numbers tell an ambivalent story. Revenue reached $13.6 billion, up 7% year over year — a positive signal. But the net loss of $3.7 billion reveals how much the restructuring is costing. The foundry division continues to burn cash, although it is attracting interest from external customers — a crucial element for the model's sustainability.

The strategic alternative is clear and whispered in Wall Street corridors: sell or spin off the foundry and become a fabless company like AMD, entrusting production to TSMC. AMD followed this path in 2009 (creating GlobalFoundries) and its stock has since grown over 5,000%.

July 2026 Update

Tan's bet is showing its first concrete results. In June 2026, three developments changed the picture:

Apple signs for 18A-P. Apple reached a preliminary agreement with Intel Foundry to manufacture some Apple Silicon processors on the 18A-P node — a high-performance variant of 18A that delivers 9% higher performance or 18% lower power consumption. It marks the first time Apple has used a non-TSMC foundry for its custom chips since 2020. First shipments are expected in Q2-Q3 2027. Apple will keep production of its flagship chips (A20) on TSMC 2nm, using Intel as a secondary partner to diversify its supply chain.

Tesla commits to 14A. Elon Musk announced that Tesla will adopt Intel's next-generation 14A node for chips in its Terafab AI complex in Austin. Google is exploring Intel's EMIB advanced packaging technology. Microsoft and Amazon are among other customers in discussions.

Yields improve. The 18A node has reached a capacity of 30,000 wafers per month across two sites (Fab 52 in Phoenix and Hillsboro, Oregon), with yields improving 7-8% monthly — in line with the industry average. The 18A-P node entered "risk production" (early production) in June 2026, with data indicating it will meet customer requirements at final qualification. Intel expects to reach industry-standard yield levels by 2027.

The fabless option has effectively been shelved. With Apple, Tesla, Google, Microsoft, and Amazon as confirmed or potential customers, Tan has met and exceeded his "2-3 significant foundry customers" target. The market responded: Intel's stock rose significantly after the Apple deal announcement.

DAMM Analysis

Delimitation (6/10): The decision is now clearer. Tan had set securing 2-3 significant foundry customers as a key criterion — a goal achieved with Apple, Tesla, and others in the pipeline. Technical metrics (yields, capacity, 14A roadmap) are measurable and progressing. The score is not higher because uncertainties remain: the Apple agreement is preliminary, 18A-P shipments are not expected until 2027, and foundry profitability has not yet been demonstrated. But the decision boundaries are much clearer than six months ago.

Asymmetry (6/10): The asymmetry has improved significantly. The upside scenario has become more probable with real customers (no longer hypothetical): Apple as a foundry client is an unprecedented technological validation for Intel. The downside has narrowed — even if 18A-P doesn't achieve full parity with TSMC, having Apple as a secondary client guarantees volumes and revenue. The catastrophic risk (foundry without customers) has been averted.

Room to Maneuver (4/10): Slightly improved. Future revenue from foundry customers (Apple, Tesla) will provide additional cash. CHIPS Act subsidies continue to flow. But Intel is still burning cash in the restructuring, and the transition to positive foundry revenue will take another 12-18 months. Room to maneuver remains tight but no longer as critical as three months ago.

Minimum Move (5/10): Tan's strategy has proven more incremental than it appeared. He first fixed the corporate culture, then achieved mass production of 18A, then launched 18A-P as a variant to attract premium customers, and is now working on 14A for the next generation. Each step built credibility for the next. The minimum move was Apple as a "pilot customer" to validate the foundry — not a multi-billion-dollar contract but a test on selected products.

DAMM Scenarios

The picture has changed radically. The question is no longer "will the foundry find customers?" but "will the foundry achieve profitability in time?".

If 18A-P yields reach competitive levels by 2027 and Apple confirms the full contract, Intel will have completed the largest turnaround in semiconductor history. If yields remain below expectations and Apple scales back its commitment, Intel will have customers but insufficient margins — an improvement over disaster, but not a victory.

Today's minimum move is to consolidate: deliver the first 18A-P chips to Apple on schedule, demonstrating that Intel Foundry can compete with TSMC not just on technology but also on execution.