Intel’s new chief executive Pat Gelsinger during the company’s Intel Unleashed: Engineering the Future virtual event made the announcement where he unveiled the company’s plans for the next step in Intel’s integrated device manufacturing (IDM) model, the IDM 2.0.
Several major manufacturing expansion plans were announced, starting with a $20 billion investment in the construction of two new factories in Arizona. The Intel CEO also disclosed the company’s plans to become a major foundry provider in the U.S. and Europe in order to serve its customers globally.
“We are setting a course for a new era of innovation and product leadership at Intel,” Gelsinger proclaimed. “Intel is the only company with the depth and breadth of software, silicon and platforms, packaging, and process with at-scale manufacturing customers can depend on for their next-generation innovations. IDM 2.0 is an elegant strategy that only Intel can deliver – and it’s a winning formula. We will use it to design the best products and manufacture them in the best way possible for every category we compete in.”, explained Gelsinger.
Gelsinger announced a four-point plan to reinvent Intel. He announced the IDM 2.0 strategy which included the formation of IDF, its extensive use of third-party foundries, a doubling down on its system on package approach and an update on it fixing its manufacturing and process technologies.
Extensive use of Third-Party Foundries
The CEO said Intel would rely more on third-party chip manufacturers than to manufacture everything on their own. This will include manufacturing some of its most groundbreaking processors, beginning in 2023. That said, he also added that the company will not be completely abandoning its history and plans for being a designer and manufacturer, and would produce most of them in-house. As a result, Intel will have the flexibility and scale necessary to optimize its roadmaps for cost, performance, schedule, and supply, giving it a competitive advantage.
Gelsinger’s commitment towards using rivals TSMC for the manufacturing of Intel’s new chipsets is an admission of the fact that chipzilla has fallen behind both in technology and capacity. It needs the help of TSMC, global foundries and Samsung to grow till it can rally things back.
$20 Billion Investment for Intel Foundry Services
Gelsinger addressed Intel’s future plans and demands with an announcement about a significant expansion of Intel’s manufacturing capability, beginning with plans for two new production lines in Arizona, located on the Ocotillo campus. It is proposed that these foundries would support the growing needs of Intel’s current products and customers, but they would also provide capacity for foundry customers. Moreover, the company foresees to speed up capital investments beyond Arizona, and Gelsinger said he plans to announce the next phase of capacity developments in the U.S., Europe and other global locations within the year.
The idea behind this is investment is to enable Intel to serve the incredible demand for semiconductor manufacturing by being a major provider of U.S.-based and European-based foundry capacity. To support Intel’s efforts in this area, a new standalone business unit, Intel Foundry Services, will be led by semiconductor industry veteran Dr Randhir Thakur. IFS will differentiate itself from other foundries by combining leading-edge process technology which Intel is championing with its new system on package approach, with renewed capacity and packaging, across the U.S. and Europe besides a world-class IP portfolio, including RISC-V/ARM cores, x86 cores and ARM ecosystem IPs for customers.
Intel for the first time in its history will allow customers to use their foundries to manufacture their semiconductors. And this isn’t just restricted to ARM or RISC-V cores but also customised x86 cores. Previously, Intel designed and manufactured x86 cores apart from the license AMD has, but now it will allow third-party customers to design these cores and manufacture the end chipset for them as a part of foundry services. This model will also be extended to customers who want ARM-based cores or RISC-V. Intel had already announced that it would open up its manufacturing facilities for ARM chipsets but the fact it has brought back veteran Sunil Shenoy as a senior VP who recently did a lot of work with the RISC-V open-source instruction set means that Intel will go big on commercialising the architecture. For ARM-based designs which are now ubiquitous for smart gadgets, self-driving technology, 5G and also gaining huge momentum in data centres and supercomputers like the Fugaku, this is another growth opportunity for Intel.
That’s why Gelsinger said despite losing Apple as a customer for the MacBook Air and MacBook Pro, he would pursue Apple as a customer as he would Apple to manufacture those M1 processors in Intel’s foundries instead of TSMC. Similarly, Gelsinger also said he would pursue Qualcomm which makes all of its Snapdragon processors at Samsung and TSMC fabs in Asia. If Apple decides to use x86 on its devices that’s a win, but even if that’s not happening, Intel can at least manufacture the chipset and still rake in a big business.
For Intel, this is also an enormous opportunity to gain more customers in the data centre space. For instance, Amazon has started making Graviton processors for AWS augmenting x86 based cores it already offers. Intel could help Amazon manufacture these chips and could also open up the potential for other cloud computing players like Microsoft Azure to make customised chipsets for their data centres based on x86 cores and manufacture them at the same time. X86 would be easier to adapt as it has been the platform for mainstream computing for decades, it was just that Intel didn’t do that bespoke work for its customers, but with the establishment of Intel Foundry Services, Intel will provide the best of its technology and IP and offer it to its customers.
System-On-Package is Intel’s lynchpin
Most modern semiconductors are SoCs or system-on-chip. Take, for instance the new M1 chip on the Apple MacBook Air. Everything is on one die manufactured on the TSMC 5 nanometer process. Intel has fallen behind in manufacturing technology, which means that its tech often doesn’t deliver the best performance per watt. But it still has many elements which are purely superior to what’s offered on ARM.
That’s why it is still the leader in traditional computing as you have a motherboard and various elements, like the GPU, CPU, RAM, storage etc are all installed on it and all of them may be manufactured on differing processes unlike an SoC but this doesn’t scale well to small gadgets or even laptops. System-On-Package or SOP is a middle ground between the fully integrated approach of SoCs and the modularity of traditional computers. Intel like AMD is betting the future will be predicated by SoPs, not SoCs which will allow differing technologies and system modules (core) to be packaged on top of each other or next to each other but packaged together in a single element.
Intel needs to fix its manufacturing issues for this to come to fruition. But this is how it intends on leapfrogging its rivals at TSMC after falling behind them in the era of the smartphone. Intel’s bet is that when the SoP wave comes, its combination of superior manufacturing technology and in-house design capability will enable it to produce the leading semiconductor stack in the world. This advantage will be exaggerated because now it is not solely betting the farm on just x86 as it will be able to deploy this manufacturing capability for ARM and RISC-V in case it gets commercialised at scale.