Find out where Intel stands in its bid to compete with fab business behemoths TSMC and Samsung.
Intel’s foray into the foundry business has been making waves for some time, and the latest episode relates to its fab expansion in Oregon. While the chipmaker has been announcing one fab project after another in the United States and abroad, what’s the actual game plan? Will Intel be able to catch up to TSMC and Samsung in foundry services? Is there a correlation between the recent chip shortage and Intel’s ambitious fab plans?
However, before we delve into answers to the above questions, here is a quick recap of the prevailing circumstances.
For a start, Intel’s move to expand the foundry industry comes at an especially opportune juncture. Over the past few years, the semiconductor universe has had a crash course in just how tenuous business operations can be when unexpected world events suddenly emerge. Now, amid the war in Ukraine, the past few weeks have only made that even clearer as international relationships have become more strained and uncertain.
The semiconductor industry has come a long way from the early days when it perpetuated the “Real men have fabs” mantra, says Laurie Balch, research director at Pedestal Research. “While we’re certainly not headed back to a full vertical integration of IC design and manufacturing, Intel’s diverse arsenal of knowledge across the development spectrum portends a shift to offering foundry customers an even greater range of expertise from their manufacturing partner.”
She expects that chip developers will benefit from access to Intel’s vast experience in semiconductor design methodologies and tools, as well as with manufacturing best practices for a wide variety of semiconductor technologies and processes. “It also presents tremendous business opportunities for Intel in offering such a breadth of services to customers,” Balch added.
Figure 1 Intel brings diverse IC manufacturing expertise along with its fab services. Source: Intel
Laurent Moll, chief operating officer at Arteris IP, added that an increase in overall fab capacity will enable an expanded volume of new semiconductor products. “That’s great news.” In other words, TSMC and Samsung need completion, which will lead to innovation, keep IC manufacturing costs in check and serve as a hedge against future supply shortages.
Intel’s fab predicament
Intel has a track record of innovation, so it’s hard to bet against the company being able to execute this plan. However, going frontal with TSMC is challenging, especially with their multi-process node dominance. TSMC has been dominant, first and foremost, because of its commitment to lead in the development of lead-edge process nodes and the fact that it has always executed. Second, and more importantly, its sole business is the foundry.
Now contrast this with Intel. One of the biggest hurdles Intel will need to overcome in terms of pure-play foundry business is assuring chip vendors that they will get priority when allocation hits. After all, Intel will be using the same fabs for manufacturing its own ICs. So, will TSMC’s biggest customers take a chance with Intel for first-source leading-edge technology designs?
Next, while Intel had been known as a semiconductor process leader and innovator through most of its history, that reputation has recently been brought into question lately. While TSMC has been flawlessly executing the process roadmap, Intel got stuck at 10 nm and has been struggling at the lower process geometries.
Ironically, Samsung has also been struggling at lower nodes. According to Korean media reports, the actual yield of Samsung’s 4-nm chips is between 20% and 30%. On the other hand, TSMC’s 5-nm chips made available in 2019 are reported to yield around 80%.
Nevertheless, Intel has vowed to deliver five nodes in four years using extreme ultraviolet lithography (EUV) to reclaim its process leadership or at least join the ranks of TSMC and Samsung. “Pat Gelsinger realizes how manufacturing capacity and expertise are critical to Intel’s future and to the industry overall,” Balch said. “The semiconductor supply chain issues of the past few years really moved this issue to the forefront.”
Intel’s fab roadmap
Randhir Thakur, president of Intel Foundry Services (IFS), told EDN that Intel’s goal in launching IFS is to create a leading end-to-end foundry offering and become one of the top foundries by the end of the decade. “This is a key part of Intel’s broader IDM 2.0 strategy.” He claims that IFS has been met with an overwhelming customer response with strong support from customers and ecosystem partners since its launch in March 2021.
Since the formation of IFS, Thakur added, Intel has built a world-class team, but we know we can’t achieve this vision on our own. “That’s why we are embracing a collaborative approach to build a thriving open ecosystem to support our customers’ innovation pipeline,” he said. “To that end, we recently launched a new initiative designed to empower our ecosystem partners and foundry customers and push chip innovation further and faster.”
Through a collaboration between Intel Capital and IFS, the company is creating a $1 billion fund to support early-stage startups and established companies building disruptive technologies for the foundry ecosystem. Initially, the fund will focus on two key strategic industry inflections: enabling modular products with an open chiplet platform and supporting design approaches that leverage multiple instruction set architectures (ISAs), with a particular focus on investments and partnerships that will help drive further adoption of RISC-V.
Regarding Intel’s game plan for catching up on smaller nodes, Thakur says that Intel has laid out a clear path to restoring a leadership position in process technology while it’s going to introduce five nodes in the next four years. Intel has just delivered the first of five nodes with the recent launch of Alder Lake. “We are confident that we can continue to deliver at this pace because we have increased funding for both people and equipment,” he said.
Figure 2 The 12th-gen Alder Lake processors are built around the 10 nm SuperFin process node. Source: Intel
Intel has invested approximately $1.5 billion in people and equipment supporting this fab roadmap. “We are also embracing the best from the industry across equipment, EDA, and research,” Thakur added. “All combined, we have confidence that we will continue to execute.”
Why Tower Semiconductor?
Intel’s acquisition of pure-play fab Tower Semiconductor provides a window into the company’s larger fab ambitions. Intel’s move to buy the Israel-based fab isn’t based on technology capability because Tower is two to three generations behind foundry process nodes. Instead, it’s Tower’s foundry infrastructure and customer service that Intel wants to infuse into its new fab business.
TSMC is successful because of its manufacturing-centric culture. That’s why it’s not a transition in which a company like Intel could hire more people and get the job done. It’s worth mentioning that Intel wanted to buy Global Foundries, but it became too expensive after the IPO.
Pedestal Research’s Balch pointed to another important dynamic in the acquisition of Tower Semiconductor. “Broadening its manufacturing abilities to encompass different semiconductor technologies, as well as diversifying its geographic footprint globally, creates tremendous business opportunity and supply chain stability for Intel.”
Figure 3 The Tower deal reflects Intel’s efforts to retarget its core competency: manufacturing. Source: Intel
Thakur provides a more detailed view of why Intel bought Tower. “Acquiring Tower significantly accelerates our path to creating an end-to-end foundry offering,” he said. “Tower complements Intel’s leading-edge technology, broad IP portfolio, and scale manufacturing with specialty technology that can service ~50% of today’s foundry TAM.”
Intel expects that, besides Tower’s customer-first approach and deep customer relationships, its wide-ranging, advanced specialty analog technology portfolio within high-growth markets will accelerate the operating scale of IFS and unlock major new opportunities in the mobile and automotive markets. “With the strength and scale of Intel combined with the foundry market operational experience of Tower, the new Intel Foundry Services will be well-positioned to capture the growing foundry TAM,” Thakur added.
The great fab game
There is a saying that no tree grows to the sky. While TSMC has been widely acknowledged for its flawless execution of lower process geometries, media reports suggest that the venerable foundry in Taiwan is facing delays in its 2 nm schedule. It’s likely to be late by five quarters.
Amid the stumbles from top foundries—TSMC and Samsung—it’s becoming more likely that Intel can catch up after it fell behind the IC manufacturing trajectory during the 2010s. And competition is always good, so even TSMC and Samsung can benefit from this highly competitive fab turf. Especially when today’s foundry opportunity is $100 billion, and it’s expected to grow significantly by the end of this decade.
Chip shortages have been widely credited to the outbreak of Covid-19, and there is a truth in that, but it’s also a fact that the emerging markets like automotive, artificial intelligence (AI), and high-performance computing (HPC) has been sharing the existing fab capacity at a higher premium. So, there is an incentive to build new fab capacity after all.
It’s important to note that while chipmakers created the bottleneck by failing to plan ahead for new fabs, the new fab build-out will take years, not months. The great fab game has just started, and a lot is yet to happen and to be seen.
This article was originally published on EDN.
Majeed Ahmad, editor-in-chief of EDN and Planet Analog, has covered the electronics design industry for more than two decades.