Building a Grand Alliance, Creating a Circle of Winners: The Triangular Dilemma of Intel, AMD, and TSMC

Before diving into the triangular dilemma between Intel, AMD, and TSMC, let’s first examine their Q4 2022 financial results.

TSMC reported a net profit of NT$295.9 billion (nearly US$10 billion), marking a staggering 77.84% year-over-year growth. The outlook for 2023 also projected continued growth.

In contrast, Intel reported a loss of US$664 million and maintained a bleak forecast for 2023.

AMD posted a modest profit of US$21 million and anticipated a recovery in its PC business starting from the second quarter of 2023.

These financial results, though brief, reflect a fundamental shift in the semiconductor industry. The rise of the foundry model and professional specialization is rapidly reshaping the landscape.

Intel has been losing market share to the “AMD + TSMC” alliance, with 2020 proving especially pivotal. That year marked the first crossover in market share between Intel and AMD. While Intel’s laptop CPUs were still produced using its 14nm process (comparable to TSMC’s 10nm), AMD leveraged TSMC’s 7nm process and captured half the desktop PC market.

Despite Intel and AMD being comparable in design capabilities, process technology had become the decisive factor—and Intel had fallen behind TSMC by more than one and a half generations, or over three years.

Beyond AMD, Intel also faced another formidable rival in Apple. In 2020, Apple—a major Intel client—announced that all its Mac products would transition to Apple-designed chips. By year-end, Apple released three Mac models using its M1 chip, delivering up to 3.5 times the performance of Intel’s previous generation.

Intel’s generational improvements had been incremental, often just 10%, whereas Apple’s M1 represented a leap of 150–200% in performance.

This breakthrough from Apple shattered Intel’s monopoly and set a trend: tech giants like Microsoft, Alphabet, Amazon, and Meta soon began designing their own chips.

Intel not only lost Apple as a key customer but was also defeated by Apple’s partnership-based model with TSMC.

Since 2020, Apple has consistently ranked as TSMC’s largest customer, accounting for 26% of TSMC’s revenue in 2021.

The “IC design customer + TSMC” model transformed the semiconductor ecosystem. And even before the PC industry shift, a more radical change had already occurred in the smartphone sector, ignited by the smartphone revolution of 2007.

In this revolution, besides IC design houses like Qualcomm and MediaTek relying on TSMC, another player gained prominence—Arm, a British company offering high-efficiency, high-performance semiconductor IP.

Founded in 1990 in Cambridge, and now majority-owned by Japan’s SoftBank, Arm’s IP is licensed in over 90% of mobile devices globally.

The model of “IC Design + IP + Foundry” enabled companies to focus on their strengths, fueling 13 years of rapid smartphone growth and ushering in the mobile internet era.

Notably, Intel had invested in mobile chip R&D as early as 1999 but failed repeatedly.

In mobile chips, where mass production and cost-cutting are vital, Intel couldn’t match TSMC’s scalability and efficiency. In 2019, Intel sold its mobile modem business to Apple.

What Intel couldn’t do, Apple succeeded in—by outsourcing production to TSMC and relying on Arm’s IP, instead of insisting on in-house manufacturing like Intel.

Although Intel has now fully exited the smartphone market, its core segments—PCs and servers—are under siege.

Qualcomm has introduced chips for “Always Connected PCs” and Chromebooks, targeting non-mainstream and entry-level segments.

Rather than using Arm’s IP, these chips are powered by IP from Nuvia, a company Qualcomm acquired in 2021, and have already been adopted by firms like Dell.

This is yet another textbook case of the “IC Design + IP + Foundry” success formula.

Fang Sung-jen（方頌仁）, a partner at DY Consulting and a former employee at Texas Instruments (TI) in the 1990s, recalls how TI’s 0.18μm process was benchmarked against two companies: “Little i” for performance and “Little t” for cost.

While losing to Intel in performance was understandable, losing to a relatively unknown Taiwanese company in cost was shocking.

“Little i” was Intel, and “Little t” was TSMC. Fang was so struck by Asia’s rising potential that he left TI and joined UMC in Taiwan.

Competing as a Team: TSMC’s Superior Model

TSMC’s rise over Intel stems from the very business model Morris Chang（張忠謀）championed.

While Intel owns and operates its fabs, it lacks product diversity. Mature processes often go underutilized, causing inefficiencies.

In contrast, TSMC and UMC leverage their mature nodes by servicing multiple clients—particularly automotive and power ICs—ensuring high utilization and strong profitability, even from fully depreciated lines.

TSMC trains its processes through high-end clients like Apple and NVIDIA, while backfilling capacity with mature-market customers.

This dual-track strategy exemplifies foundry efficiency and scalability.

As NVIDIA CEO Jensen Huang once put it, TSMC dances with 400 partners, while Intel dances alone.

In TSMC’s early days, Morris Chang approached Intel’s founders—whom he had known in his youth—for investment, but was turned down.

Back then, the IDM model dominated the world. Even TSMC’s early backer, Philips, was an IDM firm.

Philips invested primarily because of its operational presence in Taiwan and strong faith in the island’s manufacturing.

Driven by Philips Asia President Loo Yee-chiang（羅益強）’s support, the investment happened—but even they didn’t anticipate TSMC’s later success.

Philips exited entirely in 1997 to refocus on healthcare.

Lin Mao-hsiung（林茂雄）, a former head of R&D at TSMC, recalls that in the early years, TSMC struggled with fluctuating orders.

To fill excess capacity, they explored DRAM manufacturing partnerships.

In 1993, Lin and his team traveled to Siemens’ semiconductor headquarters in Munich to pitch a DRAM collaboration.

“To our surprise, the meeting lasted less than thirty minutes before we were asked to leave,” he recalled.

After a 20-hour flight from Taiwan, they returned empty-handed.

These stories show how TSMC, once repeatedly turned away by global giants, persisted step by step in a world that underestimated it—until it earned its current stature.

Ironically, Germany—the same country that dismissed TSMC—came seeking help in 2020.

Following the COVID-19 outbreak, severe shortages of automotive ICs disrupted German carmakers.

Germany’s Ministry for Economic Affairs formally requested Taiwan and TSMC to prioritize chip supply.

Lin recalled that many executives from big firms didn’t place orders with TSMC early on because they had their own fabs.

However, they stayed in touch—perhaps seeing TSMC as a future partner if they left to launch startups.

Indeed, many IC design entrepreneurs in today’s Silicon Valley were once managers TSMC had visited.

Looking back, the rise of the foundry model has become an irreversible trend in the semiconductor industry.

The contrasting trajectories of Intel, AMD, and TSMC—some declining, others ascending—serve as the most vivid and undeniable evidence of how deeply specialization has reshaped the sector.

Years ago, Morris Chang proposed the concept of the “TSMC Grand Alliance”—a term he borrowed from World War II’s Allied coalition.

Today, the alliance includes not just traditional clients like AMD, NVIDIA, Qualcomm, Broadcom, and MediaTek, but also tech giants like Apple, Microsoft, Tesla, Amazon, and Alphabet.

It also spans IP leaders like Arm, EDA firms such as Synopsys, and equipment makers including ASML.

This web of specialization and collaboration has granted TSMC transformative influence over the semiconductor industry.

So rather than saying TSMC has defeated Intel, it’s more accurate to say the TSMC Grand Alliance is now exerting its power.

And as this “circle of winners” continues to expand, those still outside may increasingly feel left behind.