Washington wants to build an American-led AI bloc, but its allies cannot run on summit declarations—or wait two years for a transformer.

The center of gravity in American AI policy is shifting from denial to alignment. For years, Washington’s principal diplomatic tool was defensive: restricting which chips and manufacturing equipment could reach its adversaries. The United States is now building a second, more ambitious layer on top of that regime—a coalition-management strategy that uses power grids, not merely semiconductors, as leverage to bind allies into a durable American-led technology bloc. The State Department gave this approach an institutional home in December 2025, when Under Secretary for Economic Affairs Jacob Helberg launched the Pax Silica initiative. Its premise is straightforward: Computational supremacy cannot be secured through semiconductor export controls alone, because frontier AI systems are ultimately constrained by the utility infrastructure that powers them. And infrastructure dependence, unlike a licensing rule, is durable and difficult to reverse.

Pax Silica now serves as a standing negotiating forum for roughly three dozen partner economies, including Japan, South Korea, the Netherlands, and the United Arab Emirates. It coordinates alignment across five domains: energy infrastructure, critical minerals, semiconductor manufacturing, computing pipelines, and logistics. Its second summit, held in June, produced a Joint Statement on AI Opportunity committing participants to coordinate investment in next-generation data centers and shared access to computing power. The document matters less for the obligations it imposes than for the recurring venue it creates—one through which Washington can keep drawing partners toward a common technological standard.

Three tools—and what each buys

It is worth distinguishing among the instruments Washington is using, because they carry very different diplomatic weight and impose very different costs.

The first is unilateral incentive: measures the United States controls entirely from its side of the table. Executive Order 14320, signed on July 23, 2025, directed the Commerce and State departments and the White House Office of Science and Technology Policy to establish an American AI Exports Program within 90 days. It explicitly tasked the Secretary of State with helping determine which “full-stack” export packages—bundling hardware, cloud services, data infrastructure, and cybersecurity—would receive federal financing and streamlined licensing. This is diplomacy by inducement: Washington sets the terms, and partners decide whether to accept them. It requires little negotiation and creates limited friction. But it also secures no binding commitment from the other side. A partner government can accept American financing today and hedge toward another supplier tomorrow.

The second is negotiated alignment: the Pax Silica model of summits, joint statements, and standards coordination. This approach is genuinely multilateral and less abrasive than coercion, but it is also diplomatically shallow by design. Joint statements rarely commit governments to much beyond continuing the conversation. The clearest example of both the model’s value and its limits is the emergency critical-minerals channel Washington established with Tokyo after Beijing tightened export restrictions on gallium, dysprosium, terbium, and yttrium. That channel is operational, designed to bypass a demonstrated Chinese chokepoint, precisely because it arose from a specific negotiated need rather than a general summit pledge. Negotiated alignment works best when it is this targeted. It works less well as a substitute for binding commitments in a field as capital-intensive as grid infrastructure.

The third is coercive alignment: legislative or regulatory pressure intended to compel a partner’s behavior rather than win it through negotiation. The clearest case is the Netherlands, where Washington has advanced the bipartisan Multilateral Alignment of Technology Controls on Hardware, or MATCH, Act. Introduced in early April, the measure would bar Dutch equipment maker ASML from selling or servicing its older lithography systems in China. The Dutch government formally objected, warning that the bill’s extraterritorial reach would damage ASML’s remaining business in China—roughly a third of its 2025 revenue—and strain a partnership Washington has already leaned on heavily since 2023. ASML shares fell on the news. Coercive alignment can produce rapid compliance, but the Dutch response illustrates the price: It turns a negotiating partner into a reluctant one, and reluctant partners look for room to hedge.

The test case: South Korea

South Korea shows how easily headline-grabbing “alliance deals” can exaggerate the diplomatic substance beneath them. Samsung Electronics and SK Hynix announced roughly $950 billion in long-term chip-supply partnerships with U.S. firms in July, a figure widely presented as evidence of deep alignment. Yet South Korean officials were explicit about what the agreements actually represented: long-term purchase commitments—“pre-orders,” in the words of one presidential adviser—not direct investment in grid infrastructure. A purchase commitment entrenches a commercial relationship. It does not oblige Seoul to co-finance transmission upgrades or accelerate permitting in the way a genuine infrastructure accord would.

The more consequential initiative sits beside the larger announcement and has attracted far less attention: SK Telecom’s plan, developed with Nvidia, to build as much as two gigawatts of AI data-center capacity in South Korea. That project will require meaningful government cooperation on transmission expansion if it is to succeed. The distinction between the two initiatives is the difference between commercial alignment and diplomatic alignment—and it is precisely the divide the Pax Silica strategy has yet to bridge.

The contradiction the strategy cannot escape

Beneath all three instruments lies a structural problem none of them has resolved. China is not only the target of U.S. and allied export controls; it is also the dominant global manufacturer of much of the grid equipment the alliance needs if it is to build quickly enough to matter. Chinese firms produced roughly 60 percent of the world’s electrical transformers in 2024 and now quote average delivery times of about nine months, compared with waits of more than two years from U.S. and allied suppliers. The disparity is so wide that Chinese transformer exports to the United States surged 182 percent year over year in early 2026, even as Washington sought to restrict Chinese involvement in allied AI infrastructure elsewhere.

This is the contradiction Pax Silica has not yet solved diplomatically. The coalition can prevail in every individual negotiation over standards and financing while partner governments, under intense domestic pressure to complete AI infrastructure on schedule, continue quietly buying Chinese equipment because nothing else arrives in time. No summit declaration can alter that arithmetic. Only a negotiated mechanism that materially shortens allied delivery times can do so.

The same logic reaches into defense

The three-part pattern also extends into defense. Under the White House’s National Security Presidential Memorandum on AI, NSPM-11, the administration is using interoperability standards as a unilateral incentive, pressing allies to adopt common software risk-management frameworks so their military hardware and autonomous systems remain compatible in joint operations. Congress has reinforced the effort through an AI Futures Steering Committee created under the National Defense Authorization Act and charged with standardizing allied cybersecurity policy.

On the negotiated side, Washington’s Partnership for Global Inclusivity on AI—run with Microsoft, Google, and Nvidia and backed by more than $100 million in computing credits—is aimed at a specific diplomatic objective: giving developing economies a route to sovereign digital capacity that does not pass through Beijing, before Chinese outreach programs reach them first. The approach recognizes that standards endure only when countries have a practical and affordable means of adopting them.

What the diplomacy still lacks

Judged as diplomacy, this is a genuine first-generation coalition strategy, and a fairly sophisticated one. It combines a standing multilateral forum, a unilateral financing apparatus, targeted negotiated alignment, a coercive track, and a layer devoted to defense interoperability. Yet it remains weighted toward the instruments Washington either controls unilaterally or applies coercively. It is comparatively light on the negotiated, binding infrastructure commitments that would give partner governments the confidence to build around American and allied suppliers over a multiyear horizon. Pax Silica, in its current form, produces joint statements. It does not yet produce enforceable agreements to build the grids on which its ambitions depend.

Closing that gap would require turning Pax Silica from a standards forum into a standing Energy-Compute Working Group with an explicit mandate to negotiate co-financed, cross-border transmission projects—not merely to endorse principles after the decisive infrastructure choices have been made elsewhere. It would mean pursuing plurilateral agreements on reciprocal fast-track licensing for small modular reactors. Deployment timelines for these reactors still vary dramatically among jurisdictions in ways that no export-control policy can correct; only regulatory diplomacy can. It would also mean treating grid capacity as an explicit and binding element of bilateral agreements with Japan, South Korea, and the Netherlands, rather than assuming it will emerge automatically from chip-supply deals. South Korea demonstrates that it will not.

Without that shift, the strategy’s greatest vulnerability is not the emergence of a rival diplomatic bloc. Beijing has offered nothing comparable to Pax Silica. The greater danger is slower and less dramatic: alignment eroding by default, one delayed transformer at a time, as partner governments quietly route around Washington’s preferred system under domestic pressures they cannot afford to wait out.