Walk into any modern car, data centre, or medical facility, and there’s one thing quietly running the show: semiconductors.
In June 2025, GlobalFoundries announced a $16 billion investment to expand its U.S. chip manufacturing capabilities, aligning with the Trump administration’s push for increased domestic production and the growing demand for AI technologies.
Once buried in hardware manuals and supply lists, these tiny components now sit at the heart of international negotiations, billion-dollar funding races, and national security agendas. The shift didn’t happen overnight. But when it hit, it changed everything.
When the World Ran Out of Chips
The year was 2021. Ford shut down several production lines. Gamers couldn’t find graphics cards. Laptop deliveries stretched out for months.
The reason? A shortage of microcontrollers, most of them built on older, 28nm nodes that power everything from car sensors to Wi-Fi routers.
It wasn’t a tech issue. It was a supply chain crisis. COVID-era planning errors, just-in-time inventory models, and overdependence on a few foundries collided at once.
The semiconductor stack had no room for error.
Why is 28nm still relevant?
While high-end chips like Apple’s M-series or NVIDIA’s AI GPUs run on 5nm or 3nm nodes, legacy nodes like 28nm power everyday tech. They’re built in older fabs and are often less profitable to expand.
Across the board, wait times soared. Some automotive-grade chips took up to 50 weeks to deliver. The entire stack – from silicon wafer suppliers to packaging units – became clogged.
Taiwan at the Centre of It All
Suddenly, global attention turned to one island.
Taiwan’s TSMC (Taiwan Semiconductor Manufacturing Company) builds almost all of the world’s most advanced chips. These include Apple’s processors, NVIDIA’s AI chips, and custom silicon for every major cloud provider. No other company comes close in terms of scale, yield, or consistency at leading-edge nodes.
What’s a node?
“Process node” (e.g. 3nm, 5nm) refers to how small the transistors on a chip are. Smaller transistors allow more power-efficient, faster chips.
TSMC doesn’t design chips, it just builds them. That pure-play model allowed it to pour resources into advanced manufacturing. But it also concentrated risk.
Taiwan lies just 130 km off China’s coast. Any disruption, whether due to conflict, natural disaster, or sanctions would ripple through every industry relying on cutting-edge chips.
The Big Five in the Chip Race
TSMC
The undisputed leader in logic chip manufacturing. Its 3nm process is now in volume production, and it’s pushing forward with 2nm trials. TSMC is also investing in advanced packaging tech like CoWoS and SoIC, which stack chips for performance gains.
NVIDIA
The poster child of the AI boom. Its H100 GPUs, used to train large language models, are built by TSMC using a 4nm node. Each chip has over 80 billion transistors and relies heavily on thermal and power efficiency – areas where manufacturing precision matters.
Intel
Once the process leader, Intel stumbled in the 2010s but is now regrouping. It’s doubling down on its own foundry business, offering cutting-edge nodes like Intel 20A (featuring RibbonFET and PowerVia) by 2025.
Samsung
While known for memory, Samsung Foundry is pushing into logic. It’s developing Gate-All-Around (GAA) transistors under its MBCFET branding and has rolled out an early 3nm node. Yield remains a challenge, but the momentum is building.
ASML
The Dutch company behind EUV (Extreme Ultraviolet) lithography machines. Without ASML’s systems, chips smaller than 7nm can’t be built. Each machine costs over $150 million and ASML is the only company that makes them.
EUV 101
EUV machines use 13.5nm light waves to etch ultrafine patterns onto wafers. This enables the dense transistor packing needed for modern chips.
China’s Parallel Play
China imports over $400 billion worth of chips each year. That’s more than it spends on oil. Yet its domestic capabilities are still catching up.
Its top foundry, SMIC, can mass-produce at 14nm and has demonstrated early 7nm without EUV. But due to US sanctions, it cannot access the EUV machines needed for next-gen nodes. This limits performance and yields.
Still, China is pouring money into chip design (via HiSilicon and others), EDA tools, and packaging. The goal isn’t immediate parity with TSMC – it’s long-term sanction resistance.
Without EUV, then what?
Chinese fabs rely on older DUV systems and “multi-patterning”, a workaround that’s more expensive and less efficient.
America’s Countermove
While US leads in chip design and software (Synopsys, Cadence), it ceded much of its manufacturing over the past two decades.
That’s changing. The CHIPS and Science Act, passed in 2022, unlocked over $52 billion in federal incentives for domestic semiconductor production.
Intel, TSMC, and Samsung are building fabs in Arizona, Texas, and Ohio. These will supply both cutting-edge and legacy chips. Meanwhile, the Department of Defense is investing in secure fabs for defence-grade chips – often built on older nodes, but with higher resilience requirements.
This is less about cost competitiveness, and more about strategic control.
Everyone Wants a Piece
The US isn’t the only one playing catch-up.
- European Union: Aims to produce 20% of the world’s chips by 2030.
- Japan: Reviving its semiconductor industry through Rapidus, a new venture with IBM to build 2nm fabs.
- India: Rolled out a $10 billion incentive scheme to promote local chip design and fabrication.
- South Korea: Still dominant in memory, it’s now investing further in foundry capabilities.
Is everyone trying to do everything?
Not quite. Most countries aren’t aiming for full-stack chip independence. The buzzword is “de-risking”, reducing over-reliance on any one geography.
The Hidden Battle: Materials and Inputs
While all eyes are on fabs, the raw materials behind chips have become another battleground.
Gallium, germanium, cobalt, photoresists, fluorinated gases – each is essential to different parts of the chipmaking process. Many are sourced or processed in China.
In 2023, China curbed exports of gallium and germanium in response to tech sanctions. The message was clear: control over materials equals leverage.
Now, governments are racing to diversify supply chains, stockpile minerals, and develop recycling technologies.
What’s Actually at Stake
Semiconductors are not just enablers of consumer tech. They underpin every strategic system – defence, space, telecom, cloud, AI, and biotech.
The countries that control chip design, manufacturing, and tooling shape the direction of innovation. They also gain leverage in trade, diplomacy, and industrial policy. Control over semiconductors is becoming a way to influence how fast other nations or companies can build what comes next.
In this sense, the chip war is less about short-term disruption and more about long-term control. It’s about setting the pace of the future.
A World Built on Silicon
No single country or company will own the entire chip stack. But where silicon is built, who designs it, and who controls the tooling will continue to define power and capability.
For enterprises, this means rethinking IT procurement and infrastructure planning. Hardware cycles will remain unpredictable. Access to the right machines and the right chips may depend on geopolitical alignment, not just purchasing power.
