This article first appeared in Forum, The Edge Malaysia Weekly on March 17, 2025 - March 23, 2025

Before the rise of fabless and pure-foundry semiconductor production models, integrated device manufacturers (IDMs) were king of the market. IDMs were vertically integrated corporations that controlled the full stack of chip design, manufacturing, packaging and sales. In the 1970s, Intel, Texas Instruments, Samsung, Hitachi and Toshiba were among the largest and most dominant IDMs, but the last three also used the chips in their own consumer and industrial products. As IDMs are both asset and capital-heavy, asset-light fabless chip designers such as Nvidia, Advanced Micro Devices (AMD) and Qualcomm soon evolved, preferring to outsource chip production to pure-play foundries like Taiwan Semiconductor Manufacturing Company (TSMC), GlobalFoundries and United Microelectronics Corp (UMC). This has allowed fabless and foundries to specialise, innovate faster and capture profits that were once concentrated within IDMs.

In 2024, the world’s top 10 IDMs generated US$536 billion (RM2.38 trillion) in global semiconductor sales. The top three — Samsung, Intel and SK Hynix — accounted for 65% of the total IDM sales. But the combined sales of leading fabless companies like Apple, Nvidia and Broadcom surpassed them by US$24 billion. While the top IDMs remain powerful players in the chip industry, their growth has been stagnant with compound annual growth rates (CAGR) of just 1%-4% between 2007 and 2024. In contrast, specialised fabless companies benefited from the artificial intelligence (AI) boom to experience significantly higher CAGR, ranging between 175 and 23% over the same period.

Although revenue growth among top IDMs has been slow, their stock market valuations showed different trends. Samsung and Sony’s market cap nearly tripled, while Intel’s market cap declined by US$64 billion over the same period. Samsung is the leading manufacturer of memory chips, particularly in the dynamic random access memory (DRAM) market, essential for high-speed storage in computers, servers and data centres. According to TechInsights, DRAM demand is projected to grow to US$100 billion, contributing to a broader US$250 billion memory market, driven by AI deep learning applications and the need for faster data transfer rates in data centres. Meanwhile, Sony is the leading supplier of image sensors for tech giant Apple’s iPhone line-up and other major high-end smartphone customers such as Samsung, Xiaomi, Oppo and Vivo.

In contrast, Intel once dominated the logic chip industry, supplying processors for most of the world’s computers. But it has lost its leadership position, gradually falling behind as its internal chip production failed to advance as quickly as its competitors. Meanwhile, its asset-heavy foundry business is not making profits. TSMC and other specialist foundries have secured fat contracts, leveraging extreme ultraviolet lithography (EUV) technology that Intel once helped fund but was slow to implement. This delay prevented Intel from manufacturing smaller and more advanced chips, allowing TSMC to become the largest foundry, attracting business from fabless giants like Apple, Nvidia and AMD.

Samsung and Sony grew by focusing on defensible niches, while Intel stretched itself thin between IDM and foundry services. In 2021, Intel launched its “IDM 2.0” strategy, to regain manufacturing leadership by increasing investments into building new and upgrading existing fabs. To offset the high costs of fab construction, Intel implemented an aggressive expansion plan, including building new fabs in the US and expanding advanced packaging facilities across global manufacturing sites, leveraging the US CHIPS Act subsidies. This aggressive push led to a significant increase of capital expenditure (capex) ratio from 26% to 45% while sales declined by 33% between 2021 and 2024. Unlike Intel, Samsung concentrated its major capex ratio on improving its niche memory chip business rather than overspending on broad expansion, except for a new fab in the US. In contrast, Sony did not spend aggressively on capex as complementary metal oxide semiconductor image sensors do not require the same level of advanced technology nodes as memory and logic chips.

As Intel has continued to invest heavily despite falling sales, the market is discounting Intel’s fabs, equipment and intellectual property (IP) due to growing competition from AMD and Advanced RISC Machines (ARM)-based processors, which erodes its pricing power and future returns. Similarly, Samsung’s assets are priced at about the same level as Intel due to a prolonged DRAM glut, which led to a significant decline in memory prices. Intel plays a crucial role in sustaining the US peripheral ecosystem, including equipment suppliers, material providers, electronic design automation software developers, talent and university research and development (R&D). If Intel were to falter without government support, it could undermine US technological leadership, endangering both economic and national security.

After taking office, President Donald Trump complained that Taiwan “took our chip business away” and warned “we want that business back”. When Trump hinted that he would not hesitate to impose tariffs of 100% on imported Taiwan chips, TSMC announced a US$100 billion plant expansion in the US. There were also rumours that TSMC would invest in Intel’s fab facilities, while Broadcom was eyeing Intel’s chip design capabilities.

Even if TSMC were to step in, Intel’s real challenge is the strategic choice of the right business model, which would depend on the choice of the next CEO who could turn the company around. Former CEO Craig Barrett has argued that the company should not be broken up, in contrast to the views of some former board members. The heart of the Intel dilemma is that the volume of its chip production is smaller than TSMC’s, which concentrated on producing ARM-based chips for mobile telephones and also specialist designed chips for customers like Nvidia and Apple. The other perceived weakness, which was one of the causes of the departure of former CEO Pat Gelsinger, is that the lead time to produce new chips is behind that of TSMC. As costs of foundries’ hardware, such as EUV lithographic machines, escalate, it becomes tougher and costlier to compete with speed and scale.

In short, the IDM business was an advantage when Intel chips had the scale and speed to reduce costs (through volume) and speed to delivery. Once management becomes bureaucratic and loses operational excellence or flexibility to deliver products to market at speed and design desired by the market, the margins get compressed and the company comes under financial pressure. Without a continuous cash flow from revenue to invest in R&D and innovation, you slip in terms of competitive advantage. Thus, as former legendary Intel CEO Andy Grove insisted, “you need to be paranoid to survive”. Once you lose strategic focus and execution excellence relative to the competitors, the heavy asset costs of the foundry business will drag the business down.

The semiconductor business, like all other businesses, is ruthless in terms of competition. There is a strategic cycle on top of the technological cycle. The leadership must be focused not only on the bottom line, but also the market trend of where technology is going. To invest heavily in R&D innovation, short-term profits may have to be sacrificed until the new product range begins to earn profits. To slow down R&D to keep short-term profits up is to sacrifice the ability to catch the next tech product wave. There is a perennial trade-off of trying different products and services through tinkering with new ideas, processes and manufacturing methods, meaning a flexible, nimble and innovative management team, and the temptation to live off legacy lines of products, by cutting costs including investing in innovation and talent.

Thus, whether the IDM model survives will depend on the ability of each generation of management to maintain speed, scale, scope and innovation. The old saying that one generation makes success, one holds it and the third blows it seems to run through many large institutions. Competitive markets are very cruel, except that today governments are more than willing to subsidise or support key industries for national strategic or security purposes. As long as the US government is keen on reshoring or onshoring the semiconductor manufacturing within American borders, including protection through tariff barriers, American foundries like Intel will continue to survive, although their ultimate form would depend on individual company circumstances.

In the global chip war, different countries will experiment with different types of semiconductor manufacturing models. The global supply chain structure of the semiconductor industry will not disappear, but it will be configured in ways where national security decisions will affect the outcomes for different players. Ultimately, each company and each nation will just have to make their own strategic choices in a world of grave geopolitical uncertainties.

Tan Sri Andrew Sheng writes on Asian global issues. Loh Peixin is a research associate at the George Town Institute of Open and Advanced Studies, Wawasan Open University. The writers are engaged in a major study of the tech industry in Penang.

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.