Share with peers:

Welcome to SKU Hero,

In this issue, we dive deep into the journey of AI chips. We examine the reasons for the ever-growing demand for AI chips and who the biggest benefactors are. We also take you through the AI chips supply chain and the geopolitical factors associated with it. 

Battle of the Chips 📱

World War II was fought with steel and aluminum. The Cold War was all about nuclear weapons and the space race. Today, the defining battle of our era is for technological supremacy, and at the very heart of this high-stakes battle between the U.S. and China are AI chips— tiny, unassuming pieces of hardware that hold immense power.

No larger than a fingernail, these chips are now the most critical resource in the global race for tech dominance. They power artificial intelligence, drive next-generation technologies, and even enable advanced weaponry. Simply put, AI chips are the backbone of modern innovation, and whoever controls their supply chain holds the keys to the future of global power and geopolitics.

But to fully grasp why these tiny chips have become so important, let’s start with the simplest question... 

What exactly are AI chips, and why do they matter?

AI has become a key part of our daily lives, transforming industries in ways we couldn’t have imagined just a few years ago. In healthcare, for example, AI is helping doctors analyze medical images in seconds, making diagnoses faster and more accurate. Electric vehicles (EVs) are another area where AI is making a difference, optimizing battery performance, managing energy efficiency, and enhancing safety features. 

And then there’s ChatGPT, which millions of people use every day. It’s powered by AI running massive models in the cloud, making advanced tools and solutions more accessible than ever.

But what’s powering all this AI magic? It is the tiny, incredibly powerful pieces of hardware called AI chips. These processors are designed specifically to handle the heavy workloads that AI requires. Unlike regular chips, like the ones on your laptop, AI chips are built to crunch huge amounts of data and run complex algorithms quickly and efficiently. 

Today, the demand for AI chips is growing fast—and it’s not hard to see why. In 2022, the global AI chip market was worth $14.9 billion. By 2032, it’s expected to skyrocket to $383 billion, with an annual growth rate of 38.2%. 

And there’s one company that has emerged as the biggest winner of the AI revolution, it’s Nvidia.

Most valuable player

"Timing is everything"—a phrase credited to Shakespeare, and more than 400 years later, it couldn’t be more relevant for Nvidia, the company at the heart of today’s AI revolution.

The rise of Large Language Models (LLMs) like OpenAI’s GPT-4, Google’s Bard, and Meta’s LLaMA have sparked an all-out AI race among Big Tech giants. But while they compete to build the smartest AI models, Nvidia is quietly fueling the race with the one thing they desperately need: hardware power.

You might wonder why tech companies are so obsessed with LLMs? The answer is simple: winning in AI means market dominance, product innovation, and shaping the next digital transformation. But here’s the catch: training models like GPT-4 isn’t easy. It requires crunching vast amounts of data and performing trillions of calculations, something that demands extremely powerful, efficient hardware.

This is where Nvidia steps in. Their GPUs — especially the A100 and H100 — are built specifically to handle the enormous workloads of training and scaling these massive AI models.

And it’s paying off big time. As Big Tech scrambles for AI supremacy, Nvidia has become a clear winner. The company now holds a staggering 80% share of the AI chip market for data centers. In Q2 2023 alone, Nvidia’s revenue doubled, skyrocketing 101% year-over-year to hit $13.5 billion, with data center sales surging 171% to $10.3 billion.

One key partnership that grabbed a lot of eyeballs in 2023 was Nvidia’s deal with Google. Together, they set out to “reinvent cloud infrastructure” for generative AI, marking a new milestone in the AI arms race. Interestingly, Nvidia reached a valuation of $1 trillion in May 2023. Today, it has tripled, making it one of just three companies with a market capitalization above $3 trillion.

However, Nvidia’s GPUs aren’t exclusive to Google. Every major tech player wants in. Take Meta, for example. In January 2024, Mark Zuckerberg announced that by year’s end, Meta’s computing infrastructure would include a staggering 350,000 H100 GPUs.

Another trend worth noting is Big Tech’s push to become self-sufficient in AI chip development. 

As Big Tech races to capture AI dominance, another critical focus has emerged: securing sustainable energy sources to power their growing infrastructure. Companies like Microsoft and Google are investing in nuclear power, including small modular reactors (SMRs), to meet the immense energy demands of AI and cloud computing. 

So, how are these highly coveted AI chips that billion-dollar companies are hunting like a treasure actually made?

Supply chain

The journey of an AI chip begins in some of the most unassuming places on Earth — deep within the ground, where the raw materials for these technological marvels are sourced. At the heart of this process lies silicon, the essential building block of every AI chip. Derived from quartz, silicon must be refined to an ultra-pure state to meet the stringent standards of semiconductor production.

One small town in the United States plays a significant role in this global story: Spruce Pine, North Carolina. Known as the "Silicon Valley of Mining," Spruce Pine is home to some of the world’s purest quartz deposits. This purity is critical to the semiconductor industry, as quartz from this region is estimated to produce 70% to 90% of the crucibles used in the global production of silicon chips. The quartz sold is valued at $10,000 per tonne, generating $300 million annually. 

Without Spruce Pine, the modern chip-making process would face a significant bottleneck.

Yet silicon is only part of the equation. AI chips also depend on rare earth metals like neodymium and tantalum, which are crucial for their conductivity, thermal stability, and overall performance. These metals come from countries such as China, Australia, and the Democratic Republic of Congo.  

12-inch silicon wafer
Image credit: Peellden, CC BY-SA 3.0 https://creativecommons.org/licenses/by-sa/3.0, via Wikimedia Commons

Once these raw materials are extracted and refined, the process moves to producing silicon wafers, which are the foundation of AI chips. 

At this stage, companies like Shin-Etsu Chemical and SUMCO, both based in Japan, dominate the market. As of 2020, Shin-Etsu Chemical held approximately 32% of the global silicon wafer market, positioning it as the leading manufacturer worldwide. Together, these companies account for 54% of global silicon wafer production. 

These companies transform raw silicon into perfectly polished wafers, ensuring the material is ready for the next stage of its transformation. These wafers are manufactured to exacting standards, often measured in nanometers, to support the advanced designs required for AI.

Next comes the highly specialized phase of chip design. This is where industry leaders like NVIDIA, AMD, Intel, and Google create blueprints for AI chips, tailoring them for specific tasks such as AI processing, parallel computing, or graphics rendering. These designs are incredibly intricate and require immense R&D investments, but the actual manufacturing is frequently outsourced to specialized fabs. 

Here’s where the story takes a fascinating turn: a single company, Taiwan Semiconductor Manufacturing Company Limited (TSMC), manufactures over 90% of the world’s most advanced chips. Let’s delve into the pivotal role TSMC plays in global chip manufacturing. 

Taiwan tales 

Today, TSMC is the undisputed leader in advanced chip manufacturing, a position that underscores its critical role in the global semiconductor supply chain. TSMC is responsible for producing more than 60% of the world’s semiconductors and over 90% of its most advanced chips, including those essential for AI, 5G, and high-performance computing. These advanced chips, with node sizes as small as 3 nanometers, are foundational to everything from smartphones and electric vehicles to AI servers and cloud infrastructure.

TSMC’s dominance didn’t happen overnight. Founded in 1987, the company pioneered the concept of a pure-play foundry—focusing exclusively on manufacturing chips for other companies rather than competing with its clients by designing its own chips. 

This strategic decision attracted key customers like Apple, NVIDIA, AMD, Qualcomm, and Broadcom, enabling TSMC to scale its operations while constantly innovating. By dedicating resources to refining fabrication technologies, TSMC has maintained a technological edge over competitors, such as Samsung and Intel, ensuring it remains the go-to manufacturer for cutting-edge semiconductors.

TSMC’s list of clients contains the elite names of the tech world. Apple alone accounts for 25% of TSMC’s revenue, relying on the company to produce the A-series and M-series chips that power iPhones, iPads, and MacBooks.

Similarly, NVIDIA depends on TSMC to fabricate the GPUs that drive AI advancements, while AMD relies on its advanced nodes for CPUs and GPUs. Together, TSMC serves over 500 customers and supports more than 12,000 different products across various industries.

TSMC’s production capacity is unmatched in terms of scale. In 2023, the company produced 15 million wafers, a significant portion of which was allocated to advanced nodes like 5nm, 4nm, and 3nm, which allow chips to deliver faster performance and improved energy efficiency.

These wafers are then transformed into billions of chips that power global technology. To put this into perspective, TSMC’s chips are found in over half of the world’s electronic devices, ranging from everyday consumer gadgets to high-tech industrial systems.

It’s no exaggeration to say that without Taiwan—a region smaller than the state of New York—the global tech ecosystem would come to a standstill. The industry's heavy reliance on a single nation and the concentration of the semiconductor supply chain in a geopolitically sensitive and disaster-prone region like Taiwan pose significant risks.

Fragile chips

Let’s start with the natural risks. Taiwan sits on the Pacific Ring of Fire, one of the most seismically active regions on the planet. Earthquakes are common, and while many are minor, it doesn’t take much to disrupt chip manufacturing. In 2021, a moderate earthquake briefly halted operations at TSMC, causing delays that rippled across industries. 

While Taiwan narrowly avoided the devastation of the 2004 Indian Ocean tsunami, it’s a stark reminder of how vulnerable the region is. For an industry so dependent on precision and continuity, even a small natural disaster could cause massive global disruptions.

But it’s not just nature testing the fragility of this supply chain. The COVID-19 pandemic laid bare how global disruptions can paralyze semiconductor production. When the pandemic hit, chip production and logistics were thrown into chaos. Lockdowns shut down factories, and demand for semiconductors outpaced supply, leading to shortages that crippled industries like automotive and consumer electronics. 

For example, the U.S. automotive industry faced significant challenges during the chip shortage, losing an estimated $210 billion in revenue in 2021 alone. Globally, the shortage resulted in the production of 11 million fewer vehicles, leaving automakers scrambling to meet demand. Major U.S. manufacturers like Ford and General Motors were forced to shut down production at several plants.

However, the most pressing threat looming over Taiwan and the global semiconductor industry comes from its close neighbor: China.

Taiwan is just 100 miles (161 kilometers) from China, and that tiny stretch of water—known as the Taiwan Strait—has become one of the most controversial geopolitical points in the world. China considers Taiwan a breakaway province and is looking to integrate it –  it hasn’t ruled out taking it back by force. 

In recent years, tensions have escalated dramatically. China is continuously conducting large-scale military exercises near Taiwan, with simulated naval blockades and airspace incursions. For instance, in 2022, after U.S. Speaker Nancy Pelosi visited Taiwan, China responded with its largest military drills ever, essentially rehearsing an invasion scenario.

But the concern isn’t just about losing access—it’s about who seizes control. If China were to take over Taiwan, it would gain unparalleled dominance over global technology markets. 

This is a nightmare scenario for countries like the U.S., which is heavily dependent on TSMC’s advanced chip production and power industries, which include defense and consumer electronics. According to a United States International Trade Commission report, about 44.2% of U.S. imports of logic chips were manufactured in Taiwan.

So, what steps is the United States taking to counter China’s rising influence and reduce its dependency on Taiwan for semiconductors and advanced AI chips?

American move 

The first strategy the US is deploying to reduce and eventually eliminate its dependence on Taiwan, South Korea, and China is bringing chip production back home. The story begins with the CHIPS and Science Act, enacted in 2022 by President Joe Biden. The CHIPs Act earmarked $52.7 billion to kickstart a new era of semiconductor innovation on American soil. 

This initiative wasn’t just a policy move; it was a vision to revive American production. With $39 billion allocated for manufacturing incentives, $13 billion for research and workforce development, and $2 billion to support legacy chip production, the U.S. set the stage for a domestic resurgence.

The response from industry giants was immediate and transformative. TSMC—the world’s leader in advanced chip production—announced a $40 billion investment to build a cutting-edge fab in Arizona, set to produce 4nm and 3nm chips. Intel joined the wave with fabs in Ohio and Arizona, committing over $60 billion, while Samsung and Micron unveiled plans for facilities in Texas and upstate New York, respectively, with investments exceeding $117 billion combined. 

Since the CHIPS Act was passed, companies have committed over $166 billion to semiconductor and electronics manufacturing projects across the United States. This has led to the launch of more than 80 new projects in 25 states, bringing the total projected investment in the industry to nearly $450 billion.

The second key strategy the U.S. adopted was implementing strict export controls. In 2022, the U.S. introduced sweeping measures to restrict the sale of advanced AI chips and semiconductor manufacturing tools to Chinese companies. These controls prohibited U.S. firms like Nvidia and AMD from selling their cutting-edge AI chips, such as the A100 and H100 GPUs, to China without special licenses. 

Additionally, companies like Applied Materials, Lam Research, and KLA, which produce essential tools for chip manufacturing, faced significant restrictions that aimed to curb China’s access to advanced semiconductor technologies.

The result? China’s ability to develop cutting-edge AI technologies and supercomputers took a significant hit. Meanwhile, China responded with an aggressive push for self-sufficiency by 2030 and allotted $1.4 trillion in investments for this. 

So, what is the current state of this tussle between these two superpowers? 

Current affairs 

On December 2, the U.S.-China semiconductor rivalry took another dramatic turn. The Biden administration, in one of its final moves, added 140 Chinese organizations to a trade blacklist. 

These companies, many tied to semiconductor advancements, now face severe restrictions, including a ban on purchasing high-bandwidth memory chips from the U.S.

China’s response was swift and calculated. It used its dominance on rare earth metals — China produces nearly 65% of the world’s rare earth metals and refines over 85% globally — to clamp down on U.S.-bound exports of critical minerals, including graphite, a resource essential for electric vehicle batteries and countless tech applications. 

The retaliation didn’t stop there. Chinese regulators opened an antitrust investigation into Nvidia, one of the U.S.’s leading GPU designers, scrutinizing a previously approved merger deal. This is a clear signal that China intends to push back against U.S. tech giants with the tools at its disposal.

These restrictions specifically target chips with designs of 7 nanometers or more advanced, which are essential for high-performance AI applications like accelerators and GPUs. The goal is clear: to slow down China’s progress in AI and other cutting-edge technologies by cutting off access to the tools it needs most. 

Conclusion

The race for AI supremacy has never been more intense. Big Tech companies are locked in a fierce battle for the advanced AI chips that power the future of innovation. While Nvidia has emerged as the biggest winner of the AI revolution, dominating the market with its cutting-edge GPUs, the landscape is beginning to shift. 

Companies like Google, Amazon, Apple, and Meta are no longer content with relying on third-party suppliers. Instead, they are pouring billions into developing their own custom AI chips — aiming to gain greater control, optimize performance, and reduce long-term costs.

However, beneath this wave of innovation lies a significant vulnerability: the fragile AI chip supply chain, which remains heavily concentrated in geopolitically sensitive regions like Taiwan. 

This reliance has turned AI chips into the focal point of a broader tech war between the U.S. and China, as both superpowers race to secure dominance. With export controls, national investments, and supply chain strategies intensifying on both sides, the outcome of this conflict will not only determine who leads the AI revolution but also reshape the global balance of technological and economic power.

Thank you for reading. We’ll see you at the next edition!