Silicon is the backbone of the world economy. The chips that drive the servers, smartphones, military systems, medical devices, and power grids all come from a relatively small number of facilities, mostly located in Taiwan, South Korea, the Netherlands, and the United States. For decades, that concentration was treated as a feature of efficient globalization. It is a strategic vulnerability for all major powers in 2026, as they all are trying to exploit it, and the ones who had the least say in its construction will be the ones hit hardest.
The semiconductor supply chain is splitting into two tiers, and it’s happening at a rapid pace. Once China’s AI chip market was dominated by NVIDIA with more than 90 percent market share, its domination is declining to around 50 percent as of early 2026, as Beijing’s “buy local” guidelines, US tariffs, and the upcoming AI Overwatch Act review period drive Chinese enterprises to seek alternatives to NVIDIA.
The 15th Five-Year Plan (2026-30) of China officially introduces the concept of technology self-reliance as a policy priority, and investment in SMIC, YMTC, Huawei Ascend, and the development of home-grown EDA tools is directed by the Chinese government, including for sub-commercial returns. China isn’t creating semiconductor capacity for profit maximization. It is building it to eliminate a strategic vulnerability. That distinction matters enormously for every country watching from the outside.
The US side of the curtain is equally active and equally confusing. In December 2025, the Trump Administration reversed Biden-era presumption of denial of Nvidia’s H200 chip to China, replacing that stance with case-by-case reviews and a 25 percent tariff per shipment to China. Trump framed it explicitly in transactional terms: “We’re allowing them to do it, but the United States is getting 25 percent of the chips, in terms of the dollar value.”
At the same time, Congress moved the AI OVERWATCH Act, which would also classify advanced semiconductor exports as weapons and ban Nvidia’s Blackwell chips from being sent to foreign entities of concern for two years. The executive branch is easing restrictions. Congress is tightening them up. The chip industry is caught in between. Every company making sourcing decisions for the next five years is operating inside that contradiction.
What Tech Sovereignty Actually Requires
The phrase “tech sovereignty” has become the most repeated term in digital policy in 2026. Worldwide investments in sovereign AI infrastructure are expected to exceed $100 billion this year. The UK has deployed a £500 million Sovereign AI Fund. Canada committed $2 billion to its Sovereign AI Compute Strategy. France committed €109 billion to its national AI plan. China currently has a $47 billion state AI development fund. The US CHIPS Act allocated $52.7 billion to domestic semiconductor manufacturing. These are the commitments of countries that already have research universities, industrial bases, engineering workforces, and capital markets capable of absorbing them. It’s a completely different discussion for a nation that doesn’t have any of those on a large scale.
The Atlantic Council’s analysis notes that international debates about digital sovereignty will overlap with AI-specific concerns about compute access, leading to proposed geographic restrictions on access to AI infrastructure and manufacturing capabilities. Geographic access restrictions on computing are not a theoretical discussion for Pakistan, Bangladesh, Nigeria, and dozens of other emerging countries. They are a practical question about whether the next generation of their technology industries gets built on US-ecosystem chips or Chinese-ecosystem chips, and whether that choice, once made through infrastructure procurement, can realistically be reversed a decade later when the political winds change.
The bifurcation creates a specific trap for emerging economies. Choosing the US ecosystem means access to Nvidia’s performance frontier, TSMC’s manufacturing precision, and ASML’s lithography equipment, but it also means operating inside an export control regime that can be tightened by executive order at any time, as the Biden-to-Trump transition demonstrated. While the Chinese ecosystem offers ever more sophisticated domestic solutions, financing for data center infrastructure from the Belt and Road initiative, and Huawei’s seamless stack of hardware and software, there are also data center solutions that the rest of the world’s capital markets, cloud platforms, and enterprise software ecosystems may not be able to interoperate with.
The new paradigm of globalized semiconductor supply chains is being replaced by a new paradigm of technological sovereignty. The coming 12-18 months will decide if the world is going towards managed coexistence or total decoupling. For a nation such as Pakistan, which is in the midst of negotiating a Pakistan-Turkey Digital Corridor, is pursuing Chinese infrastructure, is attracting diplomatic investment from the United States after the Iran mediation, and is attempting to create a local technology business with a budget of hundreds of millions, rather than hundreds of billions, of dollars, the Silicon Curtain isn’t a metaphor. It is the operational context within which all technology investment decisions for the coming 10 years will be taken. The curtain is being drawn by others. The cost of standing on the wrong side of it will be paid by everyone else.
