Western self-sufficiency in computer chips is just not going to happen

American microchip giant Intel is looking for a new CEO following Pat Gelsinger’s shock resignation. This represents more than just a corporate shake-up. It’s the end of an era in which one company could totally control a strategically vital American technology.

Under Intel’s roof is the entire process for making computer chips – from research to design to complex fabrication. For much of the late 20th century, this made the Californian company a paragon of American ingenuity.

Gelsinger has been a lifer at Intel. He rose to chief technology officer in the 2000s before leaving for a decade to run Dell’s data storage to cloud computing business, EMC.

His return as CEO in 2021 was seen as messianic. He promised the return of America’s chipset manufacturing supremacy from rivals like Taiwan Semiconductor Manufacturing Company (TSMC).

His vision involved funnelling billions of dollars into expanding chip-making factories in New Mexico and Oregon, and building plants in Ohio and Germany. To help enable this, the federal government committed US$7.9 billion (£6.2 billion) in subsidies as part of President Joe Biden’s Chips Act 2022.

Three years later, the company is in crisis. The board gave Gelsinger a choice: retire or be removed, so he chose the former.

Intel’s strategic importance

The US government has always nurtured the nation’s industry in semiconductors, the tiny chips found inside laptops and smartphones. As far back as the late 1950s it was paying 30 times the market rate for transistors for missile computers to California-based Fairchild Semiconductor, whose senior executives would later found Intel.

Semiconductors remain the lifeblood of the military, from hypersonic missiles to AI-powered defence systems. Yet the most advanced ones are predominantly fabricated by TSMC in Taiwan, including for American F-35 fighter jets.

China, of course, wants control of the island nation with which it used to be united. Whoever controls Taiwan’s semiconductor capabilities, according to a US congressional commission back in 2022, will have “the upper hand in every domain of warfare” – not to mention an industry at the heart of global commerce and society.

To complicate matters, major American semiconductor companies like Nvidia, AMD and Qualcomm own no factories, and all rely heavily for chip-making on TSMC/Taiwan, as well as Samsung/South Korea. The US has duly induced TSMC and Samsung to build plants respectively in Arizona and Texas. Yet as the last of the fully integrated US semiconductor manufacturers, no company has been more central than Intel to America’s strategy to bring chip-making back home.

The weight of history

Intel’s integrated model long made it the king of Silicon Valley, but it missed a crucial opportunity in the wake of the mobile revolution. It continued focusing on expensive, power-hungry CPUs (central processing units) for PCs and servers, failing to prioritise the lighter, more energy-efficient processors used in smartphones. It neither brought out its own chips, nor followed the advice of industry observers to mirror the TSMC model of manufacturing them for other firms.

This would have generated enough cash to be early in funding the brutally expensive research into the next generation of chip-making technologies. But Intel didn’t feel the need: its CPU manufacturing business relied on the previous state of the art, deep ultraviolet lithography (DUV). For years the company couldn’t resist the profit margins and free cashflow from continuing to focus on this older technology. Wall Street is always addicted to the cash machine, even amid diminishing technical momentum, so plenty investors supported the strategy.

Meanwhile, TSMC built up a formidable library of intellectual property (IP) to enable clients to design and order more chipsets easily. It mastered remote collaboration so that American chip designers didn’t even need to jump on Zoom calls with Taiwan. They could flesh out their latest technical requirements in TSMC’s virtual e-foundry, 24 hours a day.

Making massive volumes of chips for mobile devices enabled TSMC in the mid-2010s to invest before any rivals in the extreme ultraviolet lithography (EUV) used to manufacture today’s most powerful semiconductors. This made TSMC even more efficient, while setting a new chip-making standard that Samsung and eventually Intel would be forced to follow.

The foundering of Intel’s foundry

Gelsinger was keenly aware of the domino effect from Intel’s smartphone failure. In 2021 he launched Intel Foundry Services (IFS), a standalone unit offering TSMC-style manufacturing to third-party clients. Hence the investment in extra capacity.

Unfortunately, Intel’s corporate culture has eaten this strategy for breakfast. A great example was Intel board member Lip-Bu Tan’s resignation in August. Formerly the CEO of US chip software firm Cadence Design Systems, he had only arrived two years ago to help implement Gelsinger’s strategy.

In October 2023 he was even put in charge of manufacturing. Yet he soon quit in frustration at “the company’s lagging workforce, its approach to contract manufacturing and … risk-averse bureaucratic culture”.

His departure left a glaring gap in semiconductor expertise on the board. Intel’s stock is down 59% in 2024, and the company is cutting 15% of its workforce to save US$10 billion as IFS has struggled to take off.

Yet at bottom, this is a crisis for the US. The cherished notion of “design in America, build in America” is fading. Despite TSMC and Samsung creating US manufacturing capacity, both companies will still make the most of their products at home.

Above all, TSMC holds unparalleled chipmaking prowess and remains firmly rooted in Taiwan. Taiwan retains the key advantages in this industry: intellectual capital, skilled labour and decades of production know-how.

Meanwhile, the Taiwanese American CEO of Nvidia, Jensen Huang, whose company dominates the AI chips market, sees no reason to decouple from TSMC. And no matter how hawkish US political leaders become about overseas supply chains, economic facts persist: Tesla, for instance, relies on Nvidia’s chips, which depend on TSMC fabrication.

The global nature of chipmaking will therefore not bow to American nostalgia. The US may persuade TSMC and Samsung to open more facilities in the States, but absolute sovereignty is gone. The departure of Intel’s last true believer underscores that sobering truth.