Industrial Law Finally Receives The Deserved Respect
On June 21, Intel announced the spin-off of its foundry business, with plans to completely separate design and manufacturing and turn them into a customer-supplier relationship. This change of direction means that after a decade of hesitation, Intel’s flagship Integrated Device Manufacturer (IDM) has finally let go of the burden of combining design and manufacturing. After the spin-off, Intel-designed chips can be manufactured by TSMC (Taiwan Semiconductor Manufacturing Company) just like AMD, benefiting from TSMC’s latest manufacturing technology. This will give Intel more confidence to compete with AMD. On the other hand, Intel’s foundry, now separated from the chip design division, loses its identity as a competitor and has the opportunity to receive orders from Apple. Both sides benefit from this separation.
Even a powerhouse like Intel must bow to the laws of the industry. Domestic IDM companies in China and companies transitioning from fabless to IDM should take a lesson from this. From pure IDM to fabless+foundry is a classic evolutionary path in the semiconductor industry. From the perspective of industrial development and the trend of specialization, focusing on either fabless or foundry yields better results than being an IDM. From the broader perspective of productivity development, it is widely accepted that the finer the industrial division of labor, the higher the production efficiency. Therefore, the semiconductor industry is not exempt from the laws of industry, and the Chinese semiconductor industry is not exempt from these principles.
Intel’s Turnaround Through Spin-Off
Intel’s strategic move has paid off immediately in cost savings. According to CEO Pat Gelsinger’s plan, the spin-off is expected to save $3 billion in costs this year and contribute 6% to profits. By 2023, the foundry’s revenue is expected to exceed $20 billion, overtaking Samsung to become the world’s second-largest foundry. Cost reductions of $30 billion are expected over the next three years, with savings reaching $8-10 billion by 2025. With these cost reductions and increased efficiencies, Intel aims to catch up to TSMC by 2026 and potentially win business from major companies such as Apple and NVIDIA.
The second goal of Intel’s foundry spin-off is to avoid competing with its customers. TSMC President Mark Liu has publicly stated that the key to TSMC’s success is the trust of its customers: “Customers succeed first, and then TSMC can follow.” Because TSMC does not manufacture its products and does not compete with its customers, it has largely benefited from the industry’s significant division of labor. Without this separation, Intel would find it difficult to help its customers succeed and help them overcome their competitive identities. To address foundry customers’ concerns, Intel will establish firewalls to separate customer information and protect sensitive design data.
Initially, Intel was undecided about whether to become an Integrated Device Manufacturer (IDM) or to remain in the IDM, Fab, and Fabless models. However, with the introduction of U.S. chip legislation, which strongly supports domestic chip manufacturing, Intel was forced to take decisive action. According to the U.S. Department of Commerce’s subsidy rules, more advanced manufacturing capabilities receive greater subsidies. Intel had to make the difficult decision to spin off to catch up with advanced processes.
Insights for China
There have been many successful examples of fab and fabless separation. AMD became a fabless company by spinning off its wafer manufacturing business, which later became GlobalFoundries, one of the top ten foundries in the world. Qualcomm embraced the fabless model from its inception and grew into the industry giant it is today. To address capacity concerns, Qualcomm’s strategy included partnering with foundries to ensure flexible responses to fluctuations in market demand. IDM giants such as Infineon, NXP, and Texas Instruments have also transitioned to Fab-lite.
While these cases are well-known and the principles understood, each company faces its unique challenges. Otherwise, Intel would not be in its current situation. Domestic companies in China also have their circumstances, leading to the impulse to transition from fabless to IDM.
Domestic fabless companies have legitimate demands. On the one hand, setting up their fabs can solve capacity problems during periods of high demand for low-volume fabless products. On the other hand, in-house fabs allow for agile adaptation to rapid changes in product requirements and improve R&D efficiency from both design and fab perspectives. Therefore, the desire of these fabless companies to own fabs is understandable.
However, when the favorable factors are combined, the risks are not negligible. The construction of fabs involves substantial investments and long cycles, resulting in the dispersion of funds and resources for design companies. This can temporarily hinder the improvement of their capabilities and affect their survival during adjustment periods.
First, foundries are open platforms with advantages in technology diversification, product diversification, and a diverse customer base. As a result, the demand for foundry capacity is more stable and less susceptible to market fluctuations than that of IDMs, which are primarily self-capacity driven, have singular products, and are highly dependent on market demand and internal business conditions. Especially during periods of weak market demand, there is a high risk of idle IDM production lines. While IDM is a viable choice for foreign giants, the size of domestic design companies in China is still far from that of foreign giants, and they lack the same level of technical expertise and risk resilience.
Second, domestic IDM companies are relatively small, often with single product lines and low capacity requirements. Building a production line requires economies of scale, and many IDM companies find it difficult to recoup their investment in the production line over the years. A rough estimate is that building a 12-inch fab starts at $2 billion, with ongoing investments required for capacity expansion, R&D, and operating costs. Even established IDM giants such as Intel struggle to achieve breakthroughs in advanced process development on their own, leading them to engage in the “foundry” model.
Third, most of China’s domestic fabless companies are small, with only a few generating revenues over $1 billion. Their capacity needs are relatively small, and their inclination to move to IDM is a temporary need. Taking on too much of a fab would be like putting a big cart before a small horse. If design companies have too much control over fabs, it will create significant burdens, which may ultimately hinder the overall progress of the domestic design industry.
In conclusion, companies should not burden themselves with long-term commitments for short-term capacity needs. If there are concerns about market volatility and future capacity constraints, it would be beneficial to learn from Qualcomm’s approach to proactively securing capacity. The costs and benefits of locking in capacity and building in-house manufacturing capabilities should be carefully considered.
Conclusion
Intel, the most influential IDM leader, has undergone a complete transformation, proving once again the power of the industry division of labor. The reason there are industry rules is that they apply to everyone, regardless of time or place, without significant individual differences or lucky exceptions. Fabless companies looking to build fabs are essentially trying to address capacity issues, which are often temporary concerns during prosperous cycles, not long-term problems. Instead, companies should be more concerned about the risk of disappearing orders.
The demand for advanced foundry technology and significant investment is growing, and only neutral, platform-based, service-oriented leaders can address capacity issues. Building in-house manufacturing capacity for fabless companies is often an unrealistic romantic notion, and its drawbacks will only increase over time. Whether in the transition from IDM to fabless or from fabless to IDM, it is advisable to restrain unrealistic impulses and ultimately follow industry rules. Domestic IDM companies in China should seriously consider their future direction.
Recommended Reading:
- How Trump 2024 Election Affects the Semiconductor Sector?
- Is the Intel Era coming to an end?
- Hidden Truth: Where Are US Semiconductor Sanctions Imposed?
- The Force Behind China’s Tech: US Commerce
- Israel-Palestine Clash: How It Affects Global Chip Supply
- 2023’s Big Tech Surprise: Intel Leads Global Chip Market
- FTC Warns: Warranty Voids for Sticker Removal Illegal
