The Quiet Plan to Cut China Out of the World’s Chips

Analysis · Technology

A small ceremony took place in Washington this week, and almost no one outside the trade press noticed. Italy joined the Pax Silica initiative, the United States State Department’s signature programme for coordinating the supply chains that build the world’s most advanced computer chips.

A few days earlier, India had attended the second working session of the same body. Roughly twenty-four countries are now signed up, with more in the queue.

The membership is no longer small. It includes the United States, the United Kingdom, Japan, South Korea, Singapore, Israel, the United Arab Emirates, Australia, the Netherlands, India, Greece, Finland, Sweden, Norway, the Philippines, Qatar, the European Union, Germany and now Italy.

France is still openly skeptical, and Taiwan and Canada attended the first summit without signing.

The founding declaration does not name China. It does not need to.

The State Department’s own factsheet describes the initiative as a way to reduce “coercive dependencies” across what officials call the “full technology stack.” That stack runs from critical-mineral extraction and refining through the chemicals and tools that go into semiconductor fabrication, through the chips themselves, through the data centres that train artificial-intelligence models on those chips, all the way to the cloud services that sell the resulting intelligence to the world.

It is a remarkable amount of economic activity to coordinate quietly. The architect is Jacob Helberg, the United States Under Secretary of State for Economic Affairs, and the inaugural summit took place at what has been newly rebranded the Donald J.

Trump Institute of Peace.

That last detail is not incidental. Pax Silica is the technological wing of a broader American attempt to manage the global economy through trusted partnerships rather than the multilateral institutions of the post-1945 order.

What Pax Silica actually does

The initiative is non-binding. There is no enforcement mechanism, no treaty obligation, no formal institutional structure comparable to the World Trade Organisation or NATO.

That sounds weak. It is not.

What Pax Silica offers its members is access. The United States has, in the past three years, built a layered system of export controls on advanced semiconductor manufacturing, tooling, and computing infrastructure that would have been unimaginable in the trade-globalist atmosphere of the 1990s.

Companies inside the system get permits. Companies outside the system do not.

In early 2026 Washington also created a two-hundred-and-fifty-million-dollar seed fund to help finance the extraction and processing of critical minerals, and to seed manufacturing infrastructure in member countries. India sent senior officials to the second Pax Silica session this week with the explicit goal of accessing that fund.

Indian officials know that a fund of that size cannot, by itself, pay for the multi-billion-dollar semiconductor fabrication plants the country wants to build. But they read the fund as a deposit on a deeper relationship, the kind that opens doors at American agencies and unlocks complementary investment from American venture and corporate capital.

That is the logic of the entire initiative. Pax Silica works not by punishing non-members but by giving members preferential access to the system that controls who can build what, where, and with whose tools.

The most important non-member is Taiwan. The island makes around sixty percent of the world’s chips and roughly ninety percent of the most advanced ones, principally through TSMC, and Pax Silica without Taiwan is geometrically incomplete.

Taiwanese officials have attended the summit sessions without signing the declaration. That is the diplomatic compromise required because formal Taiwanese signature would force a confrontation with Beijing that nobody in the room wants.

The unspoken understanding is that the island participates in the architecture without lending it the political legitimacy of a signed accession. It is, in effect, a member with deniability.

Why the wall-builder is also the weak host

The unusual feature of Pax Silica is that the country it is built to contain is, by the data, in the middle of its own difficult moment. The wall-builder is having a stronger run than the host.

In April 2026, Chinese banks recorded their first contraction in new yuan-denominated lending in nine months. New loans shrank by roughly ten billion yuan, against an analyst expectation of three hundred billion yuan in fresh lending and an actual figure of nearly three trillion yuan only one month earlier in March.

For a banking system with seventy trillion dollars in assets, a contraction at that scale is not noise. It is a signal that demand for credit, even at policy rates the People’s Bank of China has cut repeatedly, has hit a soft point.

The first four months of 2026 saw eight point five nine trillion yuan in new lending, against ten point zero six trillion in the same months of 2025. Nearly fifteen percent of the credit pipeline is gone.

Where it matters more is in the composition. According to research by the Rhodium Group and Seafarer Funds, two of the most-cited analytical shops on China, state-owned enterprises with political connections continue to access bank credit at preferential rates, while private firms — both large and small — are being squeezed.

New loans to households grew at only zero point five percent year-on-year in January 2026, an all-time low. That is partly the property downturn working its way through the system, but it is also a measure of how thoroughly the private-sector confidence channel has broken.

Net interest margins for Chinese banks have fallen from around two point seven percent in 2014 to one point four percent at the end of 2025. The system is making fewer profitable loans even as Beijing repeatedly tells it to lend more.

The result is a curious geometry. Pax Silica is being built around a country whose own credit engine is running rough at exactly the moment its tools and chemicals are being structurally restricted.

China is, of course, still the second-largest economy in the world, the world’s biggest manufacturer, and a deep capability in semiconductors that has narrowed the gap with the frontier in the past three years. The enclosure is not a verdict.

But the timing is striking. The country being walled off is the same country whose private business class has, by the most credible analytical sources, stopped trusting its own state to give it room to grow.

What enclosure costs the wall-builders

The economic case against fragmentation is the case the affected industries make in public every time they are asked.

Fabrication, tooling, materials and design are too globally entangled to be cleanly separated. ASML, the Dutch maker of the lithography machines without which advanced chips cannot be made, depends on supplier networks across Germany, the United States, Japan and a long tail of niche specialists in countries that are not in Pax Silica.

TSMC’s fabrication plants run on chemicals from Japanese specialty houses, gases from American industrial-gas conglomerates, and a workforce trained substantially in the United States. Even Nvidia, the American company that has become the world’s most valuable through the artificial-intelligence chip boom, designs its products in California, manufactures them in Taiwan, packages them in South Korea or Malaysia, and sells the largest share of them to data centres operated by hyperscalers serving customers across the planet.

The Center for European Policy Analysis published a careful assessment of Pax Silica in February 2026 noting that the architecture “looks shaky.” The criteria for membership were unclear, commitments to action remained vague, and the choice of founding signatories suggested geopolitical alignment with Washington rather than the chip-supply expertise that would have made Germany, the Netherlands and France obvious central nodes.

The implicit cost of fragmenting the supply chain is that some of the friction the wall is meant to impose on the excluded country will also be imposed on the wall-builders. That cost is real.

It can be calculated, and the chip industry quietly calculates it. The estimate that circulates in the trade press is that fully separated semiconductor supply chains would cost the global economy somewhere between four hundred billion and one trillion dollars in lost efficiency.

Most of that cost would fall on consumers in the wealthy democracies that built the wall, and on the small handful of countries — chiefly Taiwan, South Korea and Japan — whose chip industries depend on serving every customer.

The counter-case, taken seriously

The honest objection is not that the enclosure is wrong, but that it cannot be built without inflicting roughly equal damage on the builders. The chip supply chain has more than fifty chokepoints where a single region controls over two-thirds of capacity, and they do not sort neatly into trusted and untrusted camps.

China still refines a dozen critical minerals and makes most of the world’s mature-node chips, so cutting it out at the frontier is feasible while cutting it out at the foundation is not.

The risk is that the wall-builders convince themselves the enclosure can be finished, spend the capital to keep building, and find the costs arriving before the benefits, landing first on their own consumers and firms. That is the steelmanned case, and it is supported by the same Chinese credit data the enclosure thesis cites as weakness.

The enclosure may simply move the contest rather than resolve it.

What Latin America has that the wall-builders do not

The region’s chip story is short but useful. Brazil makes some semiconductors at lower technology nodes, and Mexico does back-end packaging and testing for American firms relocating capacity under the nearshoring banner.

Chile and Bolivia sit on the lithium that will run the batteries that will draw the electricity that will train the artificial-intelligence models that the chips serve. None of these positions are central; all of them are useful.

Two of the region’s governments have already picked a side. Argentina and Chile joined the US-led AI-minerals bloc this week, betting their lithium and copper on the Pax Silica system.

That is the gift Pax Silica is unintentionally handing to the region. The two large economic blocs of the late twenty-twenties — the Pax Silica system and whatever China builds in answer — will both need access to Latin American minerals, Latin American labour, Latin American latitude.

Neither bloc can afford to push the region into the other’s arms. Each has a reason to be generous, and the region has a reason to listen to both.

The non-aligned option, in other words, has value again. The countries of the global south spent the 1960s and 1970s organising themselves around it, and it produced more posturing than results.

This time the underlying material reality — minerals, manufacturing, geography, demographic dividend — gives the posture a substance it did not have then. The question is whether Latin American governments will be sophisticated enough to use it, or whether they will, as so often before, sign away their leverage in exchange for short-term comfort.

That is the test the Chip Enclosure is putting in front of them. The countries that pass it will look, twenty years from now, like the strategic winners of the artificial-intelligence century.

The countries that fail it will not.

Frequently asked questions

What is Pax Silica?

A US State Department-led initiative, launched in December 2025, that coordinates semiconductor, critical-mineral and AI supply chains among ‘trusted partners.’ It is non-binding but gives members preferential access to the US-controlled chip system, and about two dozen countries have joined.

Why isn’t Taiwan a member?

Taiwan makes most of the world’s advanced chips through TSMC and attends the sessions, but it has not signed. A formal accession would force a confrontation with Beijing that no participant wants, so it takes part with deniability.

Does cutting out China actually work?

Only partly. China can be squeezed at the cutting-edge frontier, but it still dominates critical-mineral refining and mature-node chips, so a full separation is far harder and could cost the wall-builders $400 billion to $1 trillion.