Can Intel Achieve The Impossible Dream?

Product and Foundry: Compete and Cooperate

Intel CEO Pat Gelsinger has announced the company's intention to become the world's second-largest chip manufacturer by 2030.

Gelsinger’s turnaround plan involves not only launching a large number of new products itself, but also expanding its foundry services to all of its competitors. In essence, he wants to make the semiconductors for the companies from which he intends to win back market share in the product space.

Is this feasible? Is it achievable?

Under its new model, Intel seeks to manage the clear and obvious conflict of interest by creating a clear separation between its foundry services and product divisions. It is splitting into two legally distinct organizations under the Intel brand: Intel Foundry Services (IFS) and Intel's Product division. IFS will include the chip maker's technology development, supply chains, fabrication, and packaging services, while the Product division will focus on the development and licensing of client, desktop, and networking equipment. The two divisions will operate with minimal overlap, having separate sales forces, separate ERP systems to ensure arm's length transactions. It is suggested that Intel Foundry will treat the Intel Product division like any other customer, with fab allocations referred to as capacity corridors.

Intel EVP of manufacturing Keyvan Esfarjani explains, “We have implemented strict confidentiality measures to ensure that information about one customer is not shared with another, regardless of whether the customer is internal or external.”

The model will extend market-based pricing to its internal business units, offering them the same certainty and stability as external customers. This approach is expected to create efficiencies and contribute to greater profitability for the company.

Darren Grasby, executive Vice President for strategic partnerships and president of AMD EMEA, doesn’t think Intel is following the right path. He says, “think about the journey of AMD we had our own fabs many years ago and we chose to go fabless, and it was the turning point of the company that allowed us to invest those R&D dollars into the roadmap, and they're the roadmaps that are bringing that product and leading edge technology to market today."

Intel has always produced its own semiconductors by having its own in-house foundry. However, once the market leader, Intel has had a torrid time in recent years and has fallen from grace. Some blame its vertical integration model because capital is not being focused on any one thing, but is instead spread around. It was a case of becoming a jack of all trades and master of none.

Yet rather than follow the fabless model that has worked so well for the likes of AMD, it is pushing in the opposite direction. Rather than scale down fabrication capacity with a view to phasing it out, it is ramping it up in the hope that others will use it.

Notably, Samsung has achieved a degree of success offering foundry services to third parties and while also producing its own semiconductors in a separate products division. But Samsung has always been a distant second to TSMC in the foundry space, hampered by the fact that other companies would rather use an independent foundry instead of economically supporting a competitor. So the Samsung model is far from being a perfect precedent for Intel.

So, while Intel's internal foundry model is designed to offer significant business value and cost savings, success is far from being assured.

TSMC is by far the largest foundry business in the world, a market leader in every respect. It has enjoyed dominance for decades having produced semiconductors for most, if not all, fabless chip companies. Its customer base includes Apple, Qualcomm, AMD, Broadcom, NVIDIA, MediaTek, ARM, Google, Amazon, Microsoft and Sony to mention just a few. Notably, even Intel has utilized the services of TSMC for various components in the past.

Recent sabre rattling by China in relation to its territorial claim over Taiwan has many worried. If China were to take ownership and control of Taiwan, then the tech sector, including the US giants, would be reliant upon the People’s Republic. This is not a comfortable situation given the role that technology plays in national security.

To tackle this patent threat, TSMC hatched plans to create new foundries in the US and has progressed along the path to make that a reality, but it has not been an easy journey.

Among the challenges it has faced are cost overruns, with construction expenses in Arizona being notably higher compared to Taiwan, potentially quadrupling or quintupling the costs. Additionally, a shortage of skilled personnel proficient in the specialized expertise necessary for equipment installation in semiconductor-grade facilities has resulted in project delays. Competing with Intel and Samsung for skilled labour in the US market is a challenge that TSMC simply doesn’t have to face in Taiwan.

Moreover, logistical inefficiencies have arisen due to the absence of localized supply chains, as TSMC has spent over three decades cultivating a network of suppliers primarily situated in Asia, making replication in Arizona a time-consuming process. Regulatory bureaucracy has further hindered progress, with the project facing delays of at least a year due to cumbersome US regulations.

Cultural disparities have also emerged. TSMC has achieved outstanding success by implementing a rigorous work culture, characterized by lengthy working hours and a military-style management approach. This doesn’t work in the US and unions have presented a key challenge.

In short, the TSMC expansion into the US is driven primarily by geopolitical factors, raising concerns about the viability of the project from a business standpoint. This is relevant because if the largest player in the market which operates with huge economies of scale can’t make the foundry business work in the US, then is the soil sufficiently fertile for struggling Intel to succeed?

Morris Chang, the founder of TSMC was formerly the man at the helm of the US semiconductor industry when working at Texas Instruments in the 1960s. In an interview, hosted by the Brookings Institute and CSIS, two of the most prestigious DC think tanks, Morris called the United States’ current effort to bolster its domestic semiconductor industry “a very expensive exercise in futility”.

“Due to America’s lack of manufacturing talent and higher cost, no US-based semiconductor manufacturing facility can compete with a Taiwan-based facility, due to the cold logic of unit economics.” Morris Chang

Morris reminisced about the challenges encountered during the establishment of TSMC's inaugural manufacturing facility in the United States back in 1997, located in Oregon. Over the course of 25 years of operation, TSMC made concerted efforts to enhance performance and lower costs at this plant. Despite noticeable improvements in performance and its sustained profitability to date, the facility continues to grapple with persistently high costs, standing at 50% above those of a comparable facility in Taiwan.

So why does Gelsinger think that he can succeed where TSMC has struggled?

The answer is that he is betting the farm on 18A.

Intel's 18A process, short for 18 Angstroms, is a critical part of the company's strategy to regain technological leadership in the semiconductor industry. It represents a significant advancement in chip manufacturing technology and is the culmination of Intel's "five nodes in four years" strategy that was outlined in 2021.

The advanced technology of 18A holds the promise of providing backside power, referred to by Intel as PowerVia. This innovation entails supplying power to a chip's transistors from beneath, as opposed to the conventional method of delivering power from above through the fifteen to twenty metal layers present in existing semiconductor technology. The interference caused by feeding power top-down through these layers is addressed by the implementation of backside power. For Gelsinger, this development represents a transformative "hallelujah" moment for the chip industry.

Microsoft, Ericsson, Siemens, and a handful of others have already shown interest in the 18A node, but only time will tell whether Intel is able to deliver what is being promised for these new customers, and then whether it is able to acquire enough new foundry business to make the Gelsinger turnaround a reality.

“Microsoft are committed to supporting Intel's efforts to build a strong supply chain right here in the US. That's why we are so excited to work with Intel Foundry Services why we have chosen a chip design that we plan to produce on Intel's 18A process" Satya Nadella

Since Gelsinger's initial announcement in 2021 regarding Intel's decision to open its foundries to contract manufacturing, the company has made commitments to construct chip factories valued at over $100 billion across the United States and Europe. However, much of this endeavor relies heavily on substantial government subsidies provided through the US and EU microchip initiatives.

While public sector investment can kick-start the business, it won't ensure its ongoing sustainability. Intel must secure significant market share for long-term success. Yet, given the cost advantages in Taiwan and TSMC’s beneficial economies of scale, competing on price seems unlikely. This predicament suggests that Intel's best chance of success may unfortunately depend on unwelcome destabilizing actions by China towards Taiwan, a scenario no one wishes to materialize.

I am no expert in the semiconductor space, and provide this short summary of the situation as I understand it, in an attempt to resolve in my own mind whether or not Intel is an attractive investment opportunity.

The jury is out and it currently sits in my ‘too difficult’ pile, but I would appreciate any help and guidance in better understanding this opportunity if readers of this post are willing to share their thoughts.