Photo: Skórzewiak/Adobe Stock
Last month, the United States and Saudi Arabia signed the Strategic Artificial Intelligence Partnership. On the heels of a May 2025 agreement with the United Arab Emirates (UAE), the United States has approved sale of leading-edge semiconductors to state-run companies in the UAE’s G42 and Saudi Arabia’s Humain. Access to American compute power will enable both Gulf countries to export AI-enabled goods and services in sectors such as autonomous logistics, precision agriculture, medical diagnostics, and finance.
The Trump administration aims to ensure that “American AI technology continues to be the gold standard worldwide,” according to Vice President JD Vance. But these agreements miss an essential ingredient of American power: a guarantee that AI-enabled exports generated using American chips will be invoiced and settled in dollars. Giving other countries access to compute gives them the ability to export AI-enabled goods and services globally. That throws up a critical question: in which currency will they settle that trade—dollars, renminbi, or another currency?
Claims that compute is the new oil are now commonplace, but they often overlook what makes this comparison truly powerful. A country might spend $10 billion on data-center infrastructure using American chips—a one-time capital cost. Those chips can then generate $50–100 billion annually in AI-enabled exports to third countries, including China and the Global South: for example, they may contribute to the autonomous vehicle systems that end up on tens of thousands of vehicles, or they may help a drug discovery algorithm locate a new multi-billion dollar therapeutic. The currency used to invoice and settle these exports is a critical source of global influence.
Failing to deliver a solution to this problem would be tantamount to letting the gold standard collapse without a replacement. Fortunately for Americans, in 1974, U.S. Secretary of State Henry Kissinger and U.S. Treasury Secretary William Simon helped engineer the petrodollar system as a replacement for the gold standard in 1974. The United States didn’t just sell military equipment to Saudi Arabia—it anchored global energy markets to the dollar through an implicit agreement with lasting effects and provided the fiscal capacity that helped the United States win the Cold War.
Building a compute-dollar system poses different challenges than its predecessor petrodollar arrangement. Oil is a physical commodity with clear delivery points and standardized pricing. AI-enabled services—measured in compute units like FLOPs (floating-point operations) or AI tokens—are digital, distributed, and harder to track. For instance, a diagnostic AI trained in Abu Dhabi can serve a hospital in Berlin through cloud infrastructure spanning three continents. Without explicit enforcement mechanisms, tracking currency settlement for these transactions becomes impossible.
This is where the compute-dollar system must advance beyond its predecessor: by making dollar settlement an explicit, enforceable condition of chip access rather than an implicit understanding.
First, the United States should condition access to leading-edge chips on binding commitments to settle AI-enabled exports in dollars. The Commerce Department should modify its semiconductor licensing framework to require that any country receiving leading-edge AI chips commit that all AI-enabled services exported to third countries will be settled in dollars or dollar-backed stablecoins. This isn’t foreign policy overreach—it’s standard export control practice. Recent trade agreements with Malaysia, Cambodia, Ecuador, Argentina, and Thailand already require alignment with U.S. export controls, restrictions on transactions with sanctioned entities, and investment screening mechanisms. Adding currency settlement simply extends this proven framework to the most strategic technology of the century.
Second, the United States should use dollar-backed stablecoins as the settlement mechanism. This solves the enforcement problem that didn’t exist in 1974. In July, President Trump signed the GENIUS Act into law, establishing federal regulation of payment stablecoins backed one-to-one by U.S. dollars or short-term Treasuries. These stablecoins provide instant, programmable settlement with automatic transparency through distributed ledger technology. When a European hospital pays for an AI diagnostic service, the transaction settles immediately in dollar-backed stablecoins, creating verifiable records and sustained demand for Treasury securities held as reserves. Stablecoins can match the speed of China’s digital yuan while maintaining the dollar peg and Treasury backing—turning distributed AI services into an engine for dollar demand rather than a vector for yuan adoption.
Third, provide an economic security umbrella—a modern complement to the Cold War-style defense umbrella. Recent U.S. trade agreements already reflect elements of such an umbrella. Malaysia receives streamlined defense trade and preferential licensing in exchange for export control alignment and sanctions cooperation. The UAE’s $1.4 trillion investment commitment came with technology diversion protections. South Korea secured $25 billion in military purchases and nuclear submarine approval. Qatar’s $38 billion security partnership includes Al Udeid burden-sharing.
The compute-dollar system would formalize these arrangements—for instance, offering priority access to mineral reserves critical for AI infrastructure, participation in AI safety protocols, streamlined licensing for emerging applications, and protection from Chinese economic coercion. When Beijing threatens to cut off chip access to pressure policy changes, countries with U.S. compute-dollar agreements have guaranteed alternatives.
China grasps what’s at stake. Beijing is exporting its AI stack—chips, models, and infrastructure—while building alternative payment rails via its central bank digital currency to promote yuan settlement. Chinese officials explicitly describe the digital yuan as escaping the “weaponization” of dollar-based systems—meaning U.S. sanctions power.
When a European hospital’s diagnostic AI runs on Chinese chips settled in digital yuan, Beijing can threaten to cut access to influence policy. If the digital yuan becomes the settlement standard for AI services, the United States loses hundreds of billions in sustained dollar demand. Higher Treasury borrowing costs follow. The fiscal capacity needed to fund national priorities such as the pandemic response or bipartisan legislation such as the CHIPS and Science Act erodes. Network effects around the Chinese package of AI services and payment rails make reversal impossible once established.
The alternative is straightforward: the Secretaries of Treasury, Commerce, and State should ensure that all AI chip export licenses are conditioned on verifiable commitments to settle AI exports to third countries in dollars and dollar-based stablecoins. Commerce should issue guidance within 60 days requiring currency settlement plans for all frontier chip applications. Currently, agreements with Saudi Arabia and the UAE contain no such provisions. Negotiations with South Korea, Japan, and India are ongoing and should include the same.
The petrodollar system provided 50 years of monetary advantage that helped the United States win the Cold War, manage the 2007–08 financial crisis, and fund the industrial policy now rebuilding American manufacturing. The compute-dollar system offers comparable advantage—if we architect it now. The agreements currently being negotiated leave room for further detailing. They should be finalized to ensure that America’s technology edge translates to extended monetary power.
Navin Girishankar is president of the Economic Security and Technology Department at the Center for Strategic and International Studies in Washington, D.C.
