Intel hosted an event to launch Intel Foundry as a separate legal entity focused on delivering chips for the AI era, with the ability to serve both Intel products and external customers, including competitors.

At the Intel Foundry Direct Connect 2024 event in San Jose, Calif., the mantra from Intel CEO Pat Gelsinger was that Intel was ready to serve as a systems foundry for the AI era, aiming to be the world’s number two systems foundry as well as the most sustainable, resilient and secure foundry in the world. Many ecosystem partners came to talk about how they were delivering design tools for the Intel 18A process technology, the way they would use the technology, and how they were partnering with Intel to develop the future technology roadmap.

If there was a Disneyland for the semiconductor industry, this event at the San Jose McEnery Convention Center last Wednesday could easily come close to that. There was a huge air of excitement in response to a U.S. industry champion like Intel bringing together a whole ecosystem of partners and announcing first customers for its 18A process, plus its technology roadmap.

With 1,200 or so attendees, we seemed to be hearing a lot of comments about how people were pleased to see the wider ecosystem coming together to create the capabilities to enable all the AI compute that would be needed both now and in the near future. As OpenAI CEO Sam Altman emphasized in his fireside chat with Pat Gelsinger, “The world will need more AI compute—and that will require global investment which will require a lot of infrastructure.”

Technology roadmaps, customers and EDA partners

The announcements from Intel included more on its “five nodes in four years” (5N4Y) process roadmap, momentum with customers for the Intel 18A process, and IP and EDA vendors talking about how they were ready for Intel 18A designs.

The new roadmap includes evolutions for Intel 3, Intel 18A and Intel 14A process technologies. It also includes Intel 3-T optimized with through-silicon vias for 3D advanced packaging designs—which Intel said will soon reach manufacturing readiness. Jason Wang, president of UMC, also came on stage to talk about the new 12-nm nodes expected through the joint development between Intel and UMC. Intel Foundry plans a new node every two years and node evolutions along the way, giving customers a path to continuously evolve their offerings on Intel’s leading process technology.

On the customer front, Microsoft Chairman and CEO Satya Nadella said via video link on stage that Microsoft chose a chip design it plans to produce on the Intel 18A process.

Intel Foundry said it has design wins across foundry process generations, including Intel 18A, Intel 16 and Intel 3, along with significant customer volume on Intel Foundry Advanced Systems Assembly and Test (ASAT) capabilities, including advanced packaging. With advanced packaging expected to become a key part of enabling high-performance AI compute, Intel Foundry has created the new ASAT service to help customers deliver optimized packaging for the evolving heterogenous, multi-die and chiplet-based systems needed to enable this. You can find out more about Intel’s advanced packaging in the video interview below with Mark Gardner, Intel’s VP and GM of the packaging and test business group.

Intel’s EDA partners Synopsys, Cadence, Siemens and Ansys came on stage to talk about their tool qualification and IP readiness to enable foundry customers to accelerate advanced chip designs on Intel 18A, as well as across other Intel node families. We spoke to Aart de Geus, chairman of Synopsys, in a video interview to learn more about what the partnership with Intel means—in particular, the critical importance of having extensive IP libraries available. You can watch that video interview below.

Arm CEO Rene Haas also came on stage briefly as the company launched its new Arm Neoverse compute subsystem (CSS) designs. Welcoming Haas, Stuart Pann, senior VP of Intel Foundry, said, “There is no way we can be in the foundry business without partnering with Arm. Arm is the most important partner to have.”

CHIPS Act 2 on the way?

As Gelsinger proclaimed in his opening keynote that silicon was like God’s gift to humanity, U.S. Secretary of Commerce Gina Raimondo came on stage via video feed to assert, “We need to get back to work making more chips in America.”

To emphasize the administration’s commitment, she highlighted the announcement just two days before Intel’s event of the $1.5 billion of CHIPS Act investment earmarked for GlobalFoundries’ state-of-the-art fab in New York and modernization of its Vermont facility.

We want to see more manufacturing of leading-edge chips in this country,” Raimondo said. She even alluded to a second CHIPS Act, saying, “With respect to AI, we need to continue to invest in that. I suspect there will be a CHIPS Act 2, or continued investment.”

She also explained that the administrations’ policy was more than just one-off investments in manufacturing. “What CHIPS is about is a much bigger vision that just incentivizes new fabs—and Intel is this country’s champion chip company.”

Transformational operational changes at Intel

Gelsinger emphasized that Intel Foundry is to be run as a separate business and legal entity. In effect, Intel would have a product business and a foundry business, and Intel Foundry would serve both Intel product customers as well as external customers.

This involves a key attitudinal and operational model change at Intel, where it will embrace both internal needs as well as competitors. Intel’s executive VP and chief global operations officer, Keyvan Esfarjani, said at the conference, “This is the biggest transformation that Intel is going through in terms of business model.”

Gelsinger stated in his keynote that there was already $15 billion in lifetime value of deals for the foundry business. In the media Q&A session with him, when EE Times asked whether that was a relatively small ambition, he emphasized that it was a matter of balancing needs of internal customers and de-risking the access to advanced process technologies for external customers. “We’ll be building a lot of AI chips this decade,” he added.

In other words, Intel’s capacity expansion will benefit Intel products, but the transformation is that it will have the ability—with the foundry business running as a separate legal entity—to offer capacity to external customers, including competitors.

So, both capacity and advanced technology development would be essential for delivering on the AI promise. In this respect, Jason Wang, president of UMC, said on stage that the industry needed more cooperation than competition.

The last word probably has to go to Haas: “This was a big day for Arm.”

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved