Roger’s $0.02 - Intel’s Chip Foundry Bet (Patreon)

Intel CEO Patrick Gelsinger is positioning his company to be the world’s second-largest chip maker by 2030. To do so, Gelsinger is doing several key things. First, he’s splitting the company into two distinct businesses. Second, Intel will be leveraging cutting-edge fab technology that even TSMC hasn’t implemented yet. Third, the company is taking advantage of funding from the US’s CHIPS Act and its goal of re-shoring key technology supply chains. Finally, Gelsinger is collaborating with erstwhile competitors as well as long-time partners to help cement this new direction.

The product business will continue to develop and design the chips that we know Intel for today. The foundry business will focus on chip fabrication. Intel already operates chip foundries to produce its chips, however, this reorganization will now allow it to fabricate chips for other companies including ones using chip designs from erstwhile competitors like Arm. Decades ago, AMD went through a similar transition by spinning off its foundry operations which became Global Foundries, the third-largest chip maker in the world. What’s surprising is how long it’s taken Intel to come around to this decision.

In a nod to its relatively weak position, vis a vis the world’s largest chip maker, TSMC, Intel has acquired ASML’s Twinscan EXE:5000 lithography scanner. The Twinscan EXE is a High-NA EUV lithography scanner crucial to Intel’s plan to grow in the chip foundry space. Although the company has made gains using its technologies like Foveros 3D stacking and EMIB (embedded multi-die-interconnect bridge), the High-NA EUV will be capable of an 8nm resolution vs. the current smallest size of 13nm. It should be noted that current EUV lithography machines have sufficient precision to produce in the sub-3-nm process size, but it will require EUV double patterning, which adds time and complexity to the fabrication process. In contrast, the High-NA EUV will allow higher precision and smaller process sizes for future chips without resorting to double patterning, which could help improve yields per wafer while also speeding up production. High-NA EUV lithography has challenges, though. Current chip designs are too big to use the higher precision the scanner offers. New processes and designs will need to be developed to make full use of the technology. So, don’t expect Intel to make any commercial products with it yet. But it is rather telling that TSMC has paused on adopting this technology for now, while Intel sees it as a way to leapfrog past TSMC by 2025.

One of the flagship pieces of legislation passed recently is the US’s CHIPS and Science Act in 2022. The legislation includes $39 billion in incentives to set up semiconductor manufacturing in the US. These subsidies are critical to funding Intel’s foundry business expansion like its Ohio foundry project. By appealing to national security interests, and the increased employment benefits of having foundries located in the US, Gelsing is betting his appeals will help unlock more money. Bloomberg reports that the Biden administration is in talks with Intel to provide an additional $10 billion in subsidies.

The boldest facet of Gelsinger’s plan is working with former competitors. Last year, Intel inked a deal with Arm to make sure its 18A fabrication process would work with Arm designs. Earlier in the month, Intel got a deal in place to fabricate Faraday Technology’s new 64-core Arm design. With so many potential customers making use of Arm designs, Intel is swallowing its pride and moving away from its knee-jerk instinct to treat all Arm designs as a threat to its business. But it's not just Arm chips. GPU maker Nvidia is signaling it would not be opposed to using Intel’s Foundry Services to produce its much sought-after GPU, despite Intel’s recent push into graphics and AI hardware acceleration business.

As generative AI becomes a must-have core competency and not just an adjunct service, companies like Amazon, Microsoft, Google and others are scrambling to produce custom ICs for their operations. Intel has a golden opportunity to take business away from market leader TSMC as it struggles to meet demand. Furthermore, as the need for general-purpose x86 processors like Intel’s Xeon in servers diminishes, setting up a foundry business makes sense for the company. Instead of waiting on orders from other divisions of the company, the foundries can also solicit outside business to keep the factory in operation. Ultimately, this is Intel’s IBM play. Pivot your business to the new realities of a changing market or risk being left behind.