Intel unveiled its Intel Foundry business today which will count on external customers to its multi-billion-dollar chip manufacturing plants.

Before a crowd of more than 1,000 people, Pat Gelsinger, Intel CEO, said at the press event in San Jose, California, that the contract manufacturing business will propel the company into the forefront of making AI chips for the world.

Intel’s Foundry Direct Connect event featured special guests such as U.S. Commerce Secretary Gina Raimondo, Arm CEO Rene Haas and OpenAI CEO Sam Altman. The latter is hoping to raise trillions of dollars for AI chip manufacturing. On Intel’s part, this foundry move represents a bet of billions of dollars, and it’s the biggest wager that Gelsinger has made since he returned to Intel as CEO three years ago.

The Intel Foundry initiative aims to redefine technology, resilience, and sustainability, offering a comprehensive solution for chip designers looking to harness the power of AI.

“This is a key moment in the progression at the Intel foundry,” Gelsinger said in an interview with VentureBeat (full interview here). “We view the opportunity to position what we call the Intel Systems Foundry. A systems foundry, we believe, is a significant move from just being a wafer provider, to delivering a whole range of technologies – wafers, packaging, systems and software capability. With the surge of AI we see that the capabilities that Intel brings to the table are significantly greater than just the wafer foundry as that category has been defined today.”

The event showcased an extended process roadmap, unveiling Intel 14A process technology and specialized node evolutions. Additionally, Intel Foundry introduced Advanced System Assembly and Test (ASAT) capabilities to propel customers toward their AI ambitions.

Craig Orr, vice president of marketing and platforms at Intel, said in a press briefing that he was excited to have such broad support from across the industry.

Intel CEO Pat Gelsinger breaking ground on chip production.

“The reason I think there is such broad support and excitement for us across the industry is because we’re doing something different and we’re doing something big,” Orr said. “I think Intel’s IDM (integrated device manufacturer) 2.0 strategy was to bring out the best in the basic technology and bring out the best in the product groups and design and do that in a way that’s separate, but together.”

Orr added, “I think both of those things as continuing to get more competitive. It’s just that this Intel Foundry umbrella helps us to become more competitive as a foundry ourselves. It forces it forces us to be able to do really good benchmarking and be competitive across those key metrics. It also helps our customers trust us because they know that their data.”

Gelsinger said, “This is creating an unprecedented opportunity for the world’s most innovative chip designers and for Intel Foundry, the world’s first systems foundry for the AI era. Together, we can create new markets and revolutionize how the world uses technology to improve people’s lives.”

The extended process technology roadmap includes Intel 14A, a key addition to the ambitious five-nodes-in-four-years (5N4Y) plan. Normally, it takes a couple of years to make the transition from an existing manufacturing node to a new one. But Intel set up a plan to make five such leaps in four years.

“We’re unveiling our Intel Foundry brand, which is putting together all of the technology development, manufacturing and supply chain and the Intel Foundry services folks under one umbrella as a foundry to serve internal and external customers equally,” Orr said.

Orr added, “We’re not only just going to bring a basic foundry offering. We’re going to couple that with resilient and sustainable supply, and all of the systems-on-chip capabilities that we have as Intel, from our leadership doing this for many, many years. And as you go up that pyramid that you see the differentiation we bring becomes larger and larger.”

14A node
Intel is doubling down on chip manufacturing investments.

The roadmap outlines evolutions for manufacturing process technologies that are named Intel 3, Intel 18A process technologies. Noteworthy is Intel 3-T, optimized for 3D advanced packaging designs, reaching manufacturing readiness. Intel aims to regain process leadership with Intel 18A in 2025.

That means that, all things being equal when it comes to chip design, Intel’s chips will have a manufacturing advantage in terms of cost, efficiency, density and speed.

“With 18A we plan we plan to reestablish our process leadership from 2025 [onward]. And we will extend that further with 14A and this new roadmap we’ve announced,” Orr said.

The 14A node will bring the first use of high-end UV tools to build more intricate patterns on chips.

Orr said, “We have to keep advancing Moore’s Law over time. We have to give it a little boost with the packaging technologies that we have. And all of that has to be supported by a broad design ecosystem. But if you look at the rate of change for AI, it’s starting to grow exponentially. So these improvements that we’re having from Moore’s law, we have to add even more to that if we want to get to a point where we can bring AI everywhere in a sustainable fashion.”

Intel’s job is to optimize everything together into the most efficient system that it can. That’s what Intel has done when it has been at its best, like when it build all of the components needed for a more efficient laptop or an efficient server.

Industry partners
Five nodes in four years is on track at Intel.

Intel Foundry’s commitment to innovation extends beyond chip technology. The company plans to introduce a new node every two years, providing customers with a continuous evolution path on Intel’s leading process technology. The addition of Intel Foundry FCBGA 2D+ to its ASAT offerings further enhances the company’s comprehensive suite.

Ecosystem partners, including Synopsys, Cadence, Siemens, and Ansys, announced validated tools, design flows, and IP portfolios ready for Intel’s advanced packaging and Intel 18A process technologies. This collaborative effort ensures customers can accelerate chip designs with confidence. All of this is necessary for quickly getting chips out the door.

Intel will also make it easier to connect chips over interconnect pathways with other chiplets, Orr said.

“Within a die, we’re going to keep accelerating the amount of transistors you can get on a die. But we’re also going to give Moore’s Law a boost by putting more of those in the packages,” Orr said.

Gelsinger made it clear that Intel wants to work with all the major chip designers and bring more manufacturing back to the U.S. — one of the aims of the CHIPS Act.

“We’ve already been working with Amazon, for instance, as a packaging supplier. That’s an area where they want to work with us. Many companies doing their own chip designs want a broadening set of technologies. The ability for Intel to step in to be a broader supplier–clearly, we have Gina Raimondo participating in the event, reinforcing the criticality of the support from the U.S. government to build the western foundry and more manufacturing capacity on American soil,” Gelsinger told me.

He added, “The world needs a balanced and resilient supply chain. Everything relies on semiconductors. In that context, there’s only essentially two or three sources in the world for leading-edge logic technology. The fact that we have logic technology, we have systems capability, we have advanced packaging–it really does give us this AI-era systems foundry capability in a very unique way.”

Going green
Intel wants to provide the most sustainable foundry for chips.

Intel’s vision extends beyond technological advancements. The company aspires to be the industry’s most sustainable foundry, with a commitment to achieving 100% renewable electricity worldwide, net-positive water, and zero waste to landfills by 2030. Intel also aims for net-zero Scope 1 and Scope 2 GHG emissions by 2040 and net-zero upstream Scope 3 emissions by 2050.

Stuart Pann, senior vice president of Intel Foundry at Intel, highlighted the company’s systems approach, offering full-stack optimization from factory network to software. Pann said, “We are offering a world-class foundry, delivered from a resilient, more sustainable and secure source of supply, and complemented by unparalleled systems of chips capabilities.”

By 2030, the company wants to have 100% energy from renewable sources.

“We have to get more sustainable as an industry if we want to leave this world in good shape for the next generation. So Intel has always prided itself on being a leader in sustainability,” Orr said. “We really want to set the standard for sustainability in the foundry industry.”

Demanding applications
Intel is making a system foundry for the AI era.

Now, the AI workloads have gotten so large. It’s so complex with such high compute demands that you essentially have to create systems of chips that basically make a data center look like one big computer. To do that, you need to be able to put compute, which relies on the silicon and all the related IP interconnects to connect everything together, whether that’s on the chip or its chips across boards, across racks, together with memory, Orr said.

“We’re bringing all these capabilities together to solve this problem of complexity,” Orr said. “Making these systems for our customers is what makes this system foundry unique.”

Orr said that the number of chips shipped for AI applications nearly doubled last year. But it could have been even more if there weren’t constraints in the supply chain. That’s what Intel is trying to attack — to have enough chip factories so there won’t be chronic shortages of chips.

“To a great degree, Intel has been a manufacturer forever. I talked about IDM 1.0. As we announced our IDM 2.0 strategy a little over two years ago, it was really becoming a world-class fabless company and a world-class foundry,” Gelsinger said. “The unique ability we had to create more scale and efficiency by being a foundry for the industry. Obviously the AI surge created this moment in time where–what I call our advanced packaging is a capability that Intel has been a leader in forever. With that now, these advanced AI components need advanced packaging. It’s been a tail wind for our move into foundry.”

Gelsinger added, “We’ve been good in this area. Now, with the surge in AI, we’ve gotten a bit lucky as well. We’re seeing a lot of momentum there to accelerate the overall foundry customer engagement. Clearly, as we execute on our strategy, as we’re getting 18A done and laying out that road map, we’re getting more and more interest from major customers taking advantage of a western foundry.”

Those chips have to be more energy efficient as well so that they don’t melt down the polar ice caps.

“If we were to continue running at this rate of power, it would consume more power than 60% of power of the countries in the world. And so scaling things like this is not a path to get AI everywhere. If we want to bring AI everywhere, and we all know AI is going to be persuasive, we have to do something different.

Designers and manufacturers need to balance the compute, memory and interconnect and then map everything to the software.

“If you don’t, what ends up happening is you only use a small fraction of the total compute that you could otherwise use,” Orr said.