Today, Intel introduced its Intel Foundry company, which will rely on outside clients for its multibillion-dollar chip production facilities.
At the press conference in San Jose, California, in front of over a thousand people, Intel CEO Pat Gelsinger stated that the company’s contract manufacturing division will lead it to become a global leader in the production of AI chips.
Special attendees at Intel’s Foundry Direct Connect event included CEOs of Arm and OpenAI, as well as U.S. Commerce Secretary Gina Raimondo. The latter aims to raise trillions of dollars for the production of AI chips. This foundry move is a billion-dollar gamble on the part of Intel, and it’s the largest one Gelsinger has taken since rejoining the company three years ago as CEO.
To redefine technology, sustainability, and resilience, the Intel Foundry effort provides chip designers seeking to leverage AI with a holistic solution.
In an interview with VentureBeat (full interview here), Gelsinger stated, “This is a key moment in the progression at the Intel foundry.” We see the chance to establish what we refer to as the Intel Systems Foundry. We think that a systems foundry represents a big step forward from simply providing wafers to offering a full suite of technologies, including systems, packaging, software, and wafers. With the rise of artificial intelligence, it is evident that Intel offers far more capabilities than merely wafer foundries as those terms are now understood.
The presentation unveiled the Intel 14A process technology and the evolution of specialized nodes, as well as an expanded process roadmap. To help clients achieve their AI goals, Intel Foundry also unveiled Advanced System Assembly and Test (ASAT) capabilities.
During a press meeting, Intel Vice President of Marketing and Platforms Craig Orr expressed his excitement about receiving such widespread support from the industry.
“I believe the industry is overwhelmingly supportive and excited about us because we’re doing something different and significant,” Orr stated. The goal of Intel’s IDM (integrated device manufacturer) 2.0 strategy, in my opinion, was to combine the best aspects of design, product groups, and fundamental technology in a way that was distinct yet cohesive.
“I think both of those things are going to get more competitive,” Orr continued. All it does is make us more competitive as a foundry within the Intel Foundry umbrella. We are forced to be competitive across those critical parameters and capable of doing pretty good benchmarking. Our customers’ confidence in us is further enhanced by the knowledge that their data
“This is creating an unprecedented opportunity for the most creative chip designers worldwide as well as for Intel Foundry, the world’s first systems foundry for the AI era,” stated Gelsinger. By working together, we can transform how technology is used globally to enhance people’s lives and open up new markets.
The five nodes in four years (5N4Y) plan is an ambitious one, and Intel 14A is a significant addition to the extended process technology roadmap. The shift from an existing manufacturing node to a new one often takes a few years. But Intel outlined a five-year strategy to make five of these advancements.
“To serve internal and external customers equally, we’re unveiling our Intel Foundry brand, which unites all of the technology development, manufacturing, supply chain, and Intel Foundry services folks under one umbrella as a foundry,” Orr stated.
“We’re not just going to bring a basic foundry offering,” Orr continued. We will combine that with a stable and sustainable supply chain as well as all of Intel’s systems-on-chip capabilities, which come from our leadership’s many years of experience in this field. And you can see that the differentiation we bring gets bigger and bigger as you move up that pyramid.
Node 14A
Intel is increasing its investments in the production of chips.
The roadmap describes advancements in Intel 3, Intel 18A process technologies, which are manufacturing process technologies. Intel 3-T is noteworthy since it has reached manufacture readiness and is optimized for 3D advanced packaging designs. With Intel 18A, Intel hopes to reclaim process leadership by 2025.
This implies that Intel’s chips will have a manufacturing advantage in terms of cost, efficiency, density, and speed, assuming all else is equal in terms of chip design.
“We intend to restore our process leadership with 18A starting in 2025.” And with 14A and our recently unveiled new roadmap, we will expand it even further,” Orr stated.
The first use of sophisticated UV tools to create more complex designs on chips will be made possible by the 14A node.
Orr stated, “Over time, Moore’s Law must be continued to be advanced.” With the packaging technology we have available, we need to give it a little push. And a wide-ranging design ecology must underpin everything. However, the rate of advancement in AI is beginning to accelerate tremendously. Therefore, if we wish to reach a stage where we can implement AI globally in a sustainable way, we must add even more to the advancements brought about by Moore’s law.
It is Intel’s responsibility to combine all of these optimizations to create the most efficient system possible. When Intel has performed at its peak, such as when it has assembled all the parts required for a more effective laptop or server, that is what it has done.
The dedication of Intel Foundry to innovation goes beyond semiconductor technology. Every two years, the business intends to release a new node, giving consumers a continuous development path on Intel’s cutting-edge process technology. The full package offered by the company is further enhanced with the addition of Intel Foundry FCBGA 2D+ to its ASAT services.
Intel’s advanced packaging and Intel 18A process technologies are compatible with verified tools, design flows, and IP portfolios that ecosystem partners, including as Synopsys, Cadence, Siemens, and Ansys, have stated. Customers may confidently expedite semiconductor designs thanks to this cooperative effort. To get chips out the door fast, all of this is required.
According to Orr, Intel will also simplify the process of connecting chips with other chiplets via interconnect routes.
We will continue to increase the number of transistors that can be placed on a die. However, by including more of those in the packages, we’re also going to strengthen Moore’s Law,” Orr added.
One of the goals of the CHIPS Act, according to Gelsinger, is for Intel to collaborate with all of the major chip designers and bring more production back to the United States.
For example, we already serve as a packaging supplier for Amazon. They are interested in collaborating with us in that area. An expanding range of technologies is what many companies which design their own chips aim for. The ability of Intel to step in and be a more comprehensive supplier—Gina Raimondo is obviously attending the event—reinforcing the importance of the government’s support for the construction of the western foundry and further manufacturing capacity on American soil, Gelsinger informed me.
“The world needs a balanced and resilient supply chain,” he continued. All of it is dependent on semiconductors. In such case, the world’s primary sources of cutting-edge logic technology are effectively limited to two or three. We truly have this AI-era systems foundry capabilities in a very special way because of our sophisticated packaging, systems capability, and logic technology.
Changing To Green
Intel aims to offer the most environmentally friendly chip foundry possible.
Beyond merely advancing technology, Intel has bigger plans. By 2030, the company hopes to be the most environmentally friendly foundry in the industry and will have achieved 100% renewable electricity globally, net-positive water, and zero waste to landfills. Moreover, Intel wants to achieve net-zero emissions of greenhouse gases from Scope 1 and Scope 2 by 2040, as well as net-zero emissions from upstream Scope 3 by 2050.
Senior vice president Stuart Pann of Intel Foundry emphasized the company’s systems strategy, which provides full-stack optimization from software to the factory network. “We are providing a top-notch foundry, supplied from a dependable, more sustainable, and secure source of supply, and enhanced by unmatched systems of chips capabilities,” stated Pann.
The company’s goal is to use only renewable energy sources by 2030.
“If we want to leave this planet in excellent condition for the next generation, we as an industry need to become more sustainable. Thus, Orr stated, “Intel has always taken great satisfaction in being a leader in sustainability. “We genuinely want to lead the foundry industry in sustainability.”
The workloads related to AI have grown significantly. Because of its extreme complexity and high computation requirements, data centers are effectively made to resemble one massive computer through the use of chip systems. In order to accomplish that, Orr explained, you must be able to place computation, which depends on silicon and all associated IP interconnects to link everything, whether it be on the chip or its chips across boards, across racks, or along with memory.
Orr stated, “We’re combining all these skills to solve this complex problem.” “This system foundry is unique because we make these systems for our customers.”
According to Orr, last year saw almost twice as many chips shipped for AI applications. But if the supply chain hadn’t been constrained, it might have been far higher. Intel is attempting to combat this issue by having a enough number of chip plants to prevent ongoing chip shortages.
Intel has, in a sense, always been a manufacturer. I discussed IDM 1.0. Around two years ago, we unveiled our IDM 2.0 plan, and at the time, it was truly developing into a world-class foundry and fabless business, according to Gelsinger. Being an industry foundry gave us the unique potential to increase size and efficiency. The rise of AI gave rise to this particular moment in which Intel has long led the industry in what I refer to as “our advanced packaging.” Because of this, these sophisticated AI parts require sophisticated packaging. We’ve had a great start to our foundry shift.
We’ve done well here, Gelsinger continued. Now that AI is booming, we’ve also become a little lucky. There is a lot of momentum there to increase foundry customer interaction in general. It’s obvious that as we carry out our plan, complete 18A, and create that roadmap, we’re drawing more and more attention from significant clients looking to take advantage of a western foundry.
To prevent the melting of the polar ice caps, those chips also need to be more energy-efficient.
“If we kept using energy at this rate, we would use more energy than 60% of all the countries in the world. Therefore, implementing AI globally is not possible by scaling things in this way. We need to change if we want to implement AI everywhere, since as we all know, it will be persuasive.
Manufacturers and designers must balance memory, processing power, and connectivity before mapping it all to software.
Orr stated, “If you don’t, you wind up using only a small fraction of the total compute that you could use.”