In 1992, speaking in front of several fabless CEOs, AMD’s CEO, Jerry Sanders, notoriously remarked: “Now hear me and hear me well; real men have fabs!” However, by 2008, AMD had divested its manufacturing assets, becoming one of hundreds of semiconductor design firms embracing a new reality – the growing bifurcation of chip design and manufacturing.Today, this bifurcation continues to intensify. Global market intelligence firm IDC estimates that fabless revenue as a percentage of the semiconductor market is expected to reach a high of 62% in 2028, up from 30% in 2019. While foundry champions like TSMC represent a large part of this, they’re only one answer to the urgent question asked by governments and car companies alike. How do we secure chip capacity? Shocks like the COVID-19 pandemic and Russia’s invasion of Ukraine have refocused attention on the critical process of making semiconductors, forcing companies to consider how they can secure semiconductor capacity short of owning and operating a fab.Source: IDC, SEMICON West, July 2024Since Jerry Sanders’ proclamation, the risks of going fabless have drastically decreased, but they have not been eliminated entirely. Here are three models companies can implement to become less dependent on single sourcing and regain control over their supply chain at a lower cost.Option 1 – Develop a Captive Capacity CorridorMany companies have been choosing a captive capacity model to better balance risks and avoid being put on allocation. In this model, companies seek guaranteed manufacturing capacity through ownership of only part of the manufacturing process while defraying other costs to fab operators.For most companies, the supply chain risk of surrendering control over manufacturing is more than mitigated by eliminating the expense of the physical infrastructure required to make modern-day chips. According to a report from Boston Consulting Group (BCG), a brand-new greenfield fab in 2026 could carry a 10-year total cost of ownership (TCO) between $35 and $43 billion, which is 33% to 66% higher than in 2023. It’s no wonder that several companies opt to have someone else make chips for them. Source: Construction Physics, How to Build a $20 Billion Semiconductor Fab, May 3, 2024Captive capacity corridors can take many forms. For example, some companies are willing to make capital commitments for tooling early during an engagement, with the expectation of later receiving discounted wafers. This guarantees their ownership over part of the chip manufacturing process without worrying about owning the physical fab infrastructure or hiring fab operators.Such an arrangement can also be equally beneficial for foundries and integrated device manufacturers (IDMs). Companies often take full ownership of these tools over the long term, enabling them to service new customers.Option 2 – Establish a Joint-Venture Investment PartnershipInvesting in a new fab can be daunting even for major IDMs, leading to an increase in synergistic co-investment. In August 2023, TSMC announced a partnership with European IDMs, Bosch, Infineon, and NXP, to jointly construct a 300mm fab in Dresden, Germany that’s expected to cost upwards of €10 billion. In this arrangement, TSMC retains 70% ownership, and each other company has a 10% stake. A little more than a year later, NXP doubled down on its joint-venture playbook, announcing a $7.8 billion joint-venture with a different Taiwanese foundry, Vanguard International Semiconductor (VIS), to build a 300mm fab in Singapore. Partnering with proven foundries relieves potential concerns about efficacy, giving companies greater certainty that their investment will be maximized. Foundries also benefit by subsidizing construction and tooling costs for additional capacity while securing dedicated customers at the same time. Companies may also seek a capital injection without the complexity of having to share fab space with other manufacturers. Intel has been the focal point of this strategy over the last couple of years with two major announcements. First, the company signed an agreement with Brookfield Asset Management in August 2022, providing Intel with a new and expanded pool of capital for manufacturing build-outs. Under the agreement, the companies will jointly invest up to $30 billion in Intel’s manufacturing expansion at its Chandler, Arizona campus, with Intel funding 51% and Brookfield funding 49% of the total project cost. Intel will retain majority ownership and operating control of the two new leading-edge chip factories in Chandler, which will support long-term demand for Intel’s products and provide capacity for Intel Foundry Services (IFS) customers. Intel did this again recently, announcing a joint venture for Fab 34, a 300mm facility based in Leixlip, Ireland. Intel has agreed to sell a 49% stake to investment firm, Apollo Global Management, for $11 billion. Under this agreement, Intel will retain full ownership and operational control of the assets.Option 3 – Lease Third-Party Fab SpacePerhaps the least flashy option is to simply lease space at underutilized fabs. However, this path will often depend on market conditions that dictate available capacity. On the supply side, this is an appealing option for foundries and IDMs that suffer from underutilization. On the demand side, it can be a way to secure short-term capacity without the burden of long-term commitments.As part of the sale of its 200mm Dortmund wafer fab to Littelfuse, Elmos agreed to enter into a multi-year capacity-sharing arrangement, with Elmos buying defined volumes of wafers produced at the fab. Under this agreement, Elmos retains ownership of the fab while Littelfuse leases a portion of the manufacturing facility to fulfill the terms of the supply agreement. This arrangement allows Littelfuse to control its manufacturing space while Elmos maintains a presence at its headquarters.Another example occurred recently when Skorpios Technologies moved to a former Infineon 150mm fab in Temecula, California. The new facility provides Skorpios with a more modern cleanroom and space for future growth, and the company entered a 15-year lease with two 10-year options for the facility. Securing capacity is not a new problem for chip companies, but it has become more complicated as semiconductor manufacturing costs have skyrocketed. TSMC, with more than 60% of global foundry revenues, has propelled the chip industry forward and expanded capacity, enabling many companies to become fabless in the process. Despite its ubiquity, TSMC is not the full story. Plenty of companies, both fabless and IDMs, need to solve the problem of dedicated capacity in a landscape of escalated costs. Captive capacity corridors, joint ventures, and even traditional lease structures all signal that the future of securing chip capacity will likely be influenced by the creativity of corporate development and finance teams. 2025 is going to be an interesting year. I am excited to see what’s next!About Stephen M. RothrockStephen Rothrock founded ATREG in 2000 to help global advanced technology companies divest and acquire infrastructure-rich manufacturing assets, including wafer fabs (front-and back-end) as well as MEMS, solar, display, and R D facilities. Over the last 25 years, his firm has completed 40% of all global operational wafer fab sales in the semiconductor industry, totaling 60 transactions. Recent global acquisitions and dispositions have involved onsemi, Allegro Microsystems, Fujitsu, GLOBALFOUNDRIES, IBM, Infineon, Matsushita (Panasonic), Maxim, Micron, NXP, Sony, Qualcomm, Renesas, Texas Instruments, VIS, and more. Before founding ATREG, Stephen established Colliers International’s Global Corporate Services initiative and headed the company’s U.S. division based in Seattle. Before that, he worked as Director for Savills International commercial real estate brokerage in London, establishing their global corporate services platform that serves large multinationals and many leading technology companies. Stephen also served on the UK-listed property company’s international board and spent four years near Paris working for an international NGO. Stephen holds a master’s degree in political theology from the University of Hull, U.K., and a bachelor’s degree in business commerce from the University of Washington in Seattle.
