Key Takeaways From Intel Foundry Direct Connect 2025

Intel is working to return its manufacturing process to the leading position, while trying to attract fab semiconductor customers to make chips at its fabs. To this end, Intel must build more trust in the industry, not only to prove the competitiveness of its processes, but also to demonstrate that partners can manufacture chips in a familiar and mature workflow, consistent and highly reliable manner. Everything I hear today makes me believe Intel is moving towards this goal.

The keynote speech of the new CEO Tan Libai kicked off the day. Tan Libai is straightforward and concise. He outlines several challenges in Intel’s foundry services and the measures companies have taken to address these challenges and plan a successful route for Intel’s foundry services in the future. Tan Libo talked about the process that Intel foundry services are being implemented, which is more customer-centric and responds to customer feedback. However, in my opinion, what is most reassuring in Tan Libo's speech is the series of partners he invited to the stage and their speeches.

The success of Intel OEM services requires commitment from ecosystem partners

In today's advanced semiconductor market, no single entity can succeed alone—there are too many technologies and packaging innovations involved in modern multi-chip packaging devices. To this end, Tan Libai invited CEOs of almost all leading EDA tool providers.

Synopsys' Sassine Ghazi, Cadence's Anirudh Devgan, Siemens EDA's Mike Ellow and PDF/Solutions's John Kibarian all explained how their tools, IPs and/or processes are optimized for Intel's upcoming 18A processes and what they are doing for future nodes such as 18A-P, 14A and their derivatives. This firm commitment from Intel ecosystem partners should have a lot of weight; the vast majority of fable semiconductor manufacturers are familiar with these tools, which should facilitate the transition to Intel's foundry service manufacturing process.

All EDA providers also expressed their impressions of the Intel 18A, which were very positive. In fact, rumors among industry insiders have evaluated 18A for a long time. My many conversations with trusted people from the automotive, mobile and processor industries have shown that the Intel 18A is a very competitive process. That is, no new customer announcements were issued today. However, I suspect that once Intel foundry services start talking more about Intel 18A-P, there will be more announcements, which are a higher performance derivative of 18A and are more suitable for a wider range of applications. The 18A is already a competitive process, and Intel will deliver the next-generation Panther Lake processor based on the 18A later this year, but the 18A-P solves some technical problems that should make it more suitable for a more diverse architecture.

Anyone who carefully reviews Intel and its Intel Foundry Services initiatives must understand some of the basic realities related to the 18A. In addition to being a leading process with smaller feature sizes, it introduces two innovations – the Ribbon FET and PowerVia back-powered. It is difficult enough to bring new manufacturing processes to market, let alone adopting two new key technologies at the same time. To give full play to the full functions of 18A, chip designers need to make specific considerations from the beginning. Migrating existing designs or ongoing designs originally planned for existing processes to 18A is possible, but it may not benefit from the Ribbon FET and using the backside power supply may add additional costs. However, optimizing the design for 18A from the outset will offset most of the costs associated with power supply on the back by reducing density requirements and eliminating the need for certain metal layers on the front. It will also enable designers to take full advantage of Ribbon FETs and optimize power consumption, performance and area.

Of course, chip design is not achieved overnight, so future customers of Intel foundry services will need some time to deal with these issues and decide whether 18A, one of its derivatives, or future processes are best suited.

Intel 18A and 14A Process Milestones and Goals

As for the process itself, Intel has made many encouraging statements. 18A has achieved more than 95% of its targets and has identified areas of optimization to adjust processes to meet or exceed 100% of these targets in the next quarter. The Intel 18A should improve performance per watt by >15%, with a 1.3x chip density compared to the Intel 3, and the 18A-P will also increase performance per watt by about 8% and support 3D stacking.

Intel OEM Services also disclosed some data related to its future 14A processes. In the early stages, the 14A looks particularly powerful; Intel expects performance per watt to be 15-20% higher than that of 18A, density to be 1.3 times higher, and power consumption to be 25-35%. The Intel 14A will also use high numerical aperture EUV and introduce second-generation Ribbon FETs and PowerDirect. PowerDirect is an improved version of PowerVia, which enables direct contact power supply.

More was disclosed at the Intel Foundry Service Direct Connect Conference, including many advances in packaging technology that improve interconnect density and performance, but this cannot be described in detail in this report. It is safe to say that after digesting the information disclosed on the first day, my confidence in Intel's foundry services reached a new level. The innovations, technologies and tools to enable them seem to be in place. The problem now is to attract new customers, execute plans and bring new products to market. Given the company's huge investments and the political pressure to bring more advanced semiconductor manufacturing back to the United States, I think Intel Foundry Services can do that.

The above is the detailed content of Key Takeaways From Intel Foundry Direct Connect 2025. For more information, please follow other related articles on the PHP Chinese website!