Intel 18A Yields: Current Status Explained

Intel’s 18A process node is currently in early production, and while the company has not disclosed exact yield numbers, public statements and industry reports indicate that yields are improving but not yet at the high-volume manufacturing maturity needed for broad commercial release. This article explains what is known about 18A yields, the factors affecting them, and what to expect in the near term.

Background: Why 18A Matters

Intel’s 18A (18 angstrom) node is a critical part of the company’s “5 nodes in 4 years” strategy, aiming to regain process leadership by 2025. It uses two key innovations: RibbonFET (a gate-all-around transistor design) and PowerVia (backside power delivery). These technologies promise significant performance and power efficiency gains over previous nodes like Intel 7 and even competing nodes from TSMC and Samsung. The success of 18A is not just about Intel’s own products; it’s also the foundation of Intel Foundry’s ambition to win external customers, including Microsoft, which has already committed to using 18A for a custom chip.

How Yields Are Measured

In semiconductor manufacturing, “yield” refers to the percentage of functional dies per wafer. A high yield is essential for cost-effective production. There are several yield metrics:

• Defect density: The number of defects per square centimeter, often measured by test chips.
• Functional yield: The percentage of dies that pass all tests and meet performance specifications.
• Line yield: The percentage of wafers that complete the manufacturing process without being scrapped.

For a new node, yields typically start low and improve over time as the manufacturing process is refined. The industry considers a yield of around 90% or higher as good for high-volume manufacturing, but early production can be in the 50-70% range.

Current Status of 18A Yields

As of late 2024 and early 2025, Intel has not released official yield numbers for 18A. However, several data points provide insight:

• In August 2024, Intel announced that 18A had passed a key milestone: the first functional chips (including a PC client chip and a server chip) had been booted successfully. This indicates that the process is working, but yields were not disclosed.
• In October 2024, Intel’s CFO stated that 18A yields were “improving” and that the company was on track for production in 2025. Some analysts speculate that yields are currently in the 60-70% range, but this is not confirmed.
• Industry reports from SemiAnalysis and other analysts have suggested that 18A yields are behind TSMC’s N2 (2nm) node, which is expected to enter production in 2025. However, Intel’s use of RibbonFET and PowerVia could give it an advantage in the long run.

It’s important to note that “yield” can vary significantly depending on the chip size and complexity. A large server chip will have lower yields than a small mobile chip even on the same node.

Challenges Specific to 18A

18A introduces several technological firsts that make yield improvement more challenging:

• RibbonFET (GAA): This is Intel’s first use of gate-all-around transistors, which require precise etching and deposition of nanosheets. Any variation in these steps can lead to defects.
• PowerVia (BSPDN): Backside power delivery involves flipping the wafer and processing the backside, which adds complexity and risk of wafer breakage or contamination.
• EUV lithography: 18A uses EUV (extreme ultraviolet) lithography extensively. While not new, the number of EUV layers is higher than previous nodes, and each layer adds yield risk.

These challenges are not unique to Intel; TSMC and Samsung face similar issues with their own GAA and backside power efforts. However, being the first to implement both on a single node increases the risk.

Comparisons to Competitors

To put 18A yields in context, it’s helpful to compare with competing nodes:

• TSMC N2: Expected to enter production in 2025 with GAA transistors but no backside power. TSMC has a track record of ramping yields faster than Intel, but N2 is still in development.
• Samsung 2nm (SF2): Samsung is also developing GAA (they used it first in 3nm) but has struggled with yields on 3nm, which has limited adoption. SF2 is expected in 2025-2026.
• Intel 20A: This is a precursor node that also uses RibbonFET and PowerVia, but it is not intended for high-volume production; it serves as a test bed for 18A. Intel has said 20A yields are “good enough” for learning.

In general, Intel’s yield ramp has historically been slower than TSMC’s, but the company has improved in recent years. The key question is whether 18A can reach competitive yields by the time it needs to produce high-volume chips in 2025-2026.

Cost and Pricing Implications

Low yields directly increase the cost per die, which can affect pricing and margins. If 18A yields are below 70%, the cost per die could be significantly higher than a mature node. This is a concern for Intel’s foundry business, as external customers will compare costs with TSMC. However, Intel has stated that it expects 18A to be cost-competitive due to the performance advantages and the use of PowerVia, which can reduce the number of metal layers and thus some manufacturing steps.

Common Problems and Misconceptions

There are several common misconceptions about 18A yields:

• “Intel is struggling” vs. “Intel is behind”: Struggling implies a failure to meet expectations. While yields are not yet at high-volume levels, it’s normal for a node to take time to mature. Intel is not necessarily behind schedule; it’s just early in the ramp.
• “Yield numbers are public”: They are not. Intel only provides qualitative updates. Any specific numbers from analysts are estimates.
• “Low yields mean the node is a failure”: Not necessarily. Many nodes start with low yields and improve. The question is whether Intel can improve them fast enough to meet customer demand.

Frequently Asked Questions

What is a good yield for a new semiconductor node?

A yield of 90% or higher is generally considered good for high-volume manufacturing, but early production often starts at 50-70%. The target depends on the chip size and market; for large server chips, even 80% can be acceptable, while for mobile chips, yields need to be higher to keep costs down.

Why doesn’t Intel just use a simpler node?

Intel’s goal is to regain process leadership, which requires adopting advanced technologies like GAA and backside power. Using a simpler node would put it behind competitors in performance and efficiency, which is not viable for its product roadmap or foundry ambitions.

How long does it take to improve yields?

Typically, yields improve over 6-18 months as engineers refine the process. Intel has said it expects 18A to be in production in 2025, so we should see significant yield improvements by mid-2025. The pace depends on the complexity of the node and the experience of the team.

Will low yields affect Intel’s product launches?

If yields are too low, Intel may have to delay or limit product launches to avoid selling chips at a loss. However, Intel has a history of launching products on time even with moderate yields, by using higher prices or prioritizing high-end chips. The first 18A products are expected in 2025, and any delay would be significant.

What is the difference between 18A and 20A?

20A is a test node that uses the same RibbonFET and PowerVia technologies but is not intended for high-volume production. 18A is the production node, with optimizations for yield and performance. Intel has said that lessons from 20A are being applied to 18A.

What to Watch For

In the coming months, key indicators of 18A yield progress will include:

• Intel’s quarterly earnings calls, where they may provide updates on “defect density” or “yield improvements.”
• Announcements of external foundry customers, which would indicate that yields are good enough for commercial commitments.
• Third-party teardowns of early 18A products, which can reveal die size and performance, indirectly giving clues about yields.

Ultimately, the success of 18A will be judged by whether Intel can deliver competitive products on time and at reasonable cost. While yields are a critical factor, they are not the only one. The node’s performance and power efficiency, as well as Intel’s ability to integrate it into products, will determine its overall impact.

Related

• Mixed Water Temperature Formula and Examples
• Insulin Units to mL Conversion Calculator
• Low Voltage Calculator: Voltage Drop and Basics