Why Is the Us Restricting Chip Exports and What Does It Mean for Tech?

The Short Version

The US government has been systematically limiting the sale of advanced semiconductors – particularly high-performance chips used in AI training and development – to China. The restrictions target both the chips themselves and the equipment used to manufacture them. American companies like Nvidia, AMD, and Intel need export licenses to sell their most powerful products to Chinese customers, and many of those licenses are being denied or simply not applied for because the restriction is effectively a prohibition.

The logic, from Washington's perspective, is straightforward: advanced chips are dual-use technology. The same hardware that trains a large language model can optimize missile guidance systems, accelerate nuclear weapons simulation, and power surveillance infrastructure at scale. Allowing a geopolitical rival to freely access the frontier of computing capability is, in this framing, a national security issue dressed in commercial clothing.

How the Restrictions Escalated

The policy didn't arrive fully formed. It's been expanding in waves, each one tightening the screws further.

The initial round in October 2022 was already significant – it targeted advanced GPUs used for AI workloads, cutting Nvidia off from selling its most capable data center chips to China without a license. But there were workarounds. Nvidia responded by designing slightly downgraded chips – the A800 and H800 – that technically fell below the performance thresholds specified in the rules while still being genuinely useful for AI development. The government closed that loophole in October 2023 with a revised rule set that used a more sophisticated performance metric to catch chips that had been engineered around the previous thresholds.

The 2023 updates also extended restrictions geographically, applying similar controls to over 40 additional countries that had been considered potential re-export risks – places where chips could theoretically be purchased and then moved into China. That geographic expansion was notable because it signaled the US wasn't just managing a bilateral US-China issue but trying to control the global flow of advanced compute capacity.

In 2025, the Biden-era framework was succeeded by what the Trump administration called the "AI Diffusion Rule," restructuring the global landscape into a tiered system. A small group of close allies – roughly 18 countries including the UK, Japan, South Korea, and most of Western Europe – face essentially no restrictions. A large middle tier of countries faces caps on how much computing power they can import without a government-to-government agreement. And a group of countries including China, Russia, and several others face near-total restrictions on advanced chip imports.

What's Actually Being Restricted

To understand the stakes, it helps to be specific about what's on the restricted list. This isn't about the chips in your laptop or your phone. Consumer-grade semiconductors are largely unaffected. The restrictions target high-bandwidth, high-performance chips used specifically for AI training and inference at scale – products like Nvidia's H100 and H200 data center GPUs, and the next-generation Blackwell architecture chips. These are the workhorses of large-scale AI development, the hardware that makes it possible to train models with hundreds of billions of parameters.

Also restricted is semiconductor manufacturing equipment – specifically the extreme ultraviolet (EUV) lithography machines made by the Dutch company ASML, which are essentially the only way to manufacture the world's most advanced chips. The US has pressured the Dutch government and ASML directly to limit exports of these machines to China, which it has largely done. Without EUV equipment, fabricating chips at the leading edge (3nm and below) is currently impossible. This is arguably the more durable restriction, because chips can be redesigned or smuggled with some effort, but you can't easily replicate a machine that costs $380 million, weighs 180 tons, and requires components from dozens of specialized suppliers across multiple countries.

China's Response

Being cut off from the global semiconductor supply chain is not a problem China is accepting passively. The response has been a massive, state-directed push toward domestic semiconductor self-sufficiency that represents one of the largest industrial policy efforts in modern history.

Huawei's semiconductor subsidiary HiSilicon has continued developing advanced chips despite being on the US entity list since 2019. In 2023, Huawei surprised observers by releasing the Mate 60 Pro smartphone running a chip – the Kirin 9000S – that appeared to be manufactured at 7nm by SMIC, China's largest chipmaker, without EUV equipment. That achievement, accomplished through a technique called multi-patterning that stacks older lithography steps to achieve finer feature sizes, was widely interpreted as a signal that China's chip industry had progressed further than US policymakers assumed.

The Chinese government has committed enormous resources to domestic chip development through initiatives like the National Integrated Circuit Industry Investment Fund, colloquially known as "Big Fund," which has channeled hundreds of billions of yuan into domestic semiconductor companies. Progress is real but uneven – China can now manufacture chips at 7nm with significant effort and yield challenges, but the frontier is currently at 3nm and moving toward 2nm, and that gap represents years of development lead time that's hard to close without the equipment the restrictions are specifically designed to deny.

Meanwhile, Chinese tech companies have been stockpiling chips. In the period between each round of export control announcements, there have been documented surges in chip purchases by Chinese companies looking to build inventories before the next restriction takes effect. Nvidia's China revenue was substantial right up until the restrictions made it impossible to maintain.

What It Means for Nvidia and the US Chip Industry

The commercial impact on American semiconductor companies is real and complicated. China has historically been one of Nvidia's largest markets. The restrictions effectively closed off a significant portion of the addressable market for the company's highest-margin products. Nvidia has responded by continuing to design China-specific chips that attempt to comply with the rules while remaining useful – the H20 chip being the most recent example – but these products operate in a legal gray zone that keeps getting scrutinized and revised.

More broadly, the restrictions create a structural tension for US chip companies. They're being asked to sacrifice near-term commercial revenue in the name of national security goals that benefit the country as a whole but fall unevenly on specific companies. The semiconductor industry has pushed back on this, arguing that restricting sales doesn't actually prevent China from developing AI capability – it just means they develop it more slowly with domestic chips, while American companies lose revenue that funds the R&D needed to stay at the frontier. It's a genuine argument, not just lobbying talking points, and policymakers haven't fully resolved the tension.

The downstream effects extend beyond Nvidia. ASML, which is Dutch, has seen its China revenue affected by pressure from both the US and Dutch governments. TSMC, the Taiwanese foundry that manufactures chips for essentially everyone, has been required to stop manufacturing advanced chips for Chinese customers. The entire global supply chain has been restructured around a set of political decisions made in Washington.

The Broader Geopolitical Calculus

To understand why the US is willing to absorb commercial costs in this area, you need to understand the strategic logic that's driving the policy.

The argument runs roughly as follows: advanced AI capability will be a central determinant of military, economic, and geopolitical power over the coming decades. The country or bloc that leads in AI development will have structural advantages in autonomous weapons systems, cyber operations, economic productivity, scientific research, and intelligence gathering. Given that assessment, allowing a strategic competitor to freely access the hardware needed to close the gap – or take the lead – is not commercially neutral. It's a security risk with a price tag measured in strategic position rather than dollars.

The chip restrictions are the most visible element of a broader technology competition that also includes restrictions on advanced software, investment screening for AI and semiconductor companies, attempts to build allied semiconductor manufacturing capacity (the CHIPS Act in the US, similar legislation in Europe and Japan), and ongoing pressure on allied governments to align their own export control regimes with Washington's.

Whether this strategy will actually achieve its goals is genuinely uncertain. Export controls have a mixed historical track record – the Soviet Union developed nuclear weapons despite US efforts to restrict relevant technology, though the restrictions did slow the timeline. The hope isn't necessarily that China never develops advanced AI capability, but that the US and its allies maintain enough of a lead that the strategic balance remains favorable.

What It Means for the Rest of the Tech Industry

If you're not a chip company or a Chinese tech giant, you might be wondering where you fit into this story. The answer is: closer than you might think.

The tiered system for countries outside the US-China bilateral means that cloud computing providers, AI research institutions, and data centers in a large swath of the world now operate under new constraints about how much advanced computing capacity they can access or build. A country in the middle tier of the AI Diffusion Rule framework faces limits on the aggregate computing power it can import – limits that become binding as AI infrastructure scales up. The practical effect is that the geography of AI development is being shaped not just by investment decisions and talent but by government policy.

For developers and companies building on AI infrastructure, the restrictions affect which chips cloud providers can deploy and where, potentially influencing latency, availability, and cost of AI services in different regions. For the semiconductor industry as a whole, the restrictions are accelerating a bifurcation – a world where the US and allied ecosystems run on one set of chips and supply chains, and China increasingly runs on another. That's a massive structural change in an industry that has operated on global integration for decades.

The push for domestic manufacturing in the US through the CHIPS Act – which allocated $52.7 billion for semiconductor research and production in America – is a direct response to the realization that concentrating advanced manufacturing in a small number of locations (primarily Taiwan and South Korea) creates strategic vulnerability. Whether those investments will meaningfully shift where advanced chips are made is a question that will play out over the next decade.

FAQ

Can Chinese companies still get Nvidia chips at all? The most advanced chips – H100, H200, Blackwell-generation products – are effectively off the table without an export license that's unlikely to be granted. Nvidia has designed a China-specific chip (the H20) that complies with current restrictions while still offering meaningful performance, but even that chip has faced scrutiny and additional restrictions. The situation continues to evolve.

Does this affect consumer tech products like iPhones? Mostly no. The restrictions target data center and AI training chips, not the application processors in consumer devices. Apple designs its own chips (the A-series and M-series) which are manufactured by TSMC, and those products aren't meaningfully affected by the current export control framework.

What happens if a company violates the restrictions? Violations carry serious consequences – civil and criminal penalties, loss of export privileges, and potential placement on the Entity List, which effectively cuts a company off from US suppliers entirely. Several cases of alleged violation, including suspected chip smuggling through third-party intermediaries, have been investigated or prosecuted.

Will the restrictions actually slow China's AI development? Almost certainly yes – but by how much is genuinely uncertain. The restrictions impose real costs and slow access to the frontier of AI hardware. But China has a large talent pool, significant investment, and demonstrated ability to make progress under constraint. The restrictions likely widen the gap rather than close the door entirely.

Is this just a US-China issue? The US-China dimension is central, but the restrictions have been designed with a global architecture in mind. Allied countries are being coordinated into alignment, third-party countries are being classified into tiers, and the equipment restrictions involve Dutch and Japanese companies. This is a coalition effort more than a unilateral US policy, though Washington is clearly the driving force.

The chip export restrictions are one of those policy developments that feel dry until you realize they're rerouting the entire trajectory of the technology industry. They're reshaping where advanced chips get made, who can access them, how AI infrastructure gets built globally, and what the competitive landscape looks like between the world's two largest economies. Whether you find the policy wise or counterproductive, it's one of the most consequential tech stories of the decade – and it's just getting started.