From petrodollar to AI dollar: Can the greenback continue to rule?

[SINGAPORE] The foundations of dollar power are shifting. One of its pillars has been the petrodollar. The oil shocks of the 1970s generated enormous current-account surpluses for oil exporters, who accumulated dollar deposits and US securities while Western banks recycled some of these petrodollars as loans to developing countries.

The petrodollar’s heyday ran from the 1970s to the early 2010s, although oil revenues weakened during parts of the late 1980s and 1990s.

From 2014, its importance gradually diminished. The shale revolution reduced US dependence on imported oil, oil prices became less consistently high, Gulf states diversified their investments, and China, Russia and countries under Western sanctions experimented with non-dollar settlement. At last week’s Brics summit, members committed to promoting the use of local currencies for trade settlements.

Nevertheless, the dollar remains the dominant currency in international oil trading.

The petrodollar was more than an invoicing convention. It linked an essential commodity, dollar settlement, US financial markets and American security relationships with Gulf countries in a mutually reinforcing system.

Now, with artificial intelligence sweeping the world, some analysts foresee the rise of an “AI dollar”.

From barrels to computing power

The global economy is becoming more digital, computational and AI-intensive. The dollar could retain its pre-eminence by becoming the main currency in which AI is purchased, financed and traded. Countries and companies increasingly need dollars to buy computing power, advanced chips, cloud capacity and AI services – areas in which US companies are dominant.

The US has the pole position across the AI stack.

Nvidia, AMD and other American firms are leaders in advanced chip design. Amazon, Microsoft and Google dominate much of the international cloud market. Anthropic, Google and OpenAI operate leading proprietary models and applications. US capital markets provide much of the equity, debt and venture financing for the AI build-out. International semiconductor and computing contracts are also commonly priced in dollars.

The potential AI-dollar cycle would work like this: Foreign businesses and governments obtain dollars to buy chips, cloud capacity and AI services. American technology companies earn revenue globally. Expectations of AI-driven profits attract additional foreign capital into US equities and private markets.

Dollar stablecoins could strengthen this cycle. They represent 99.4 per cent of the market value of fiat-backed stablecoins, according to the Bank for International Settlements. If AI agents eventually use them in transactions autonomously, demand for AI services could create additional demand for digital dollars. Stablecoin issuers, in turn, generally invest their reserves in dollar deposits and short-term Treasuries.

If AI models, cloud contracts, computing tokens and autonomous transactions are routinely priced in dollars, companies will have an incentive to keep dollar balances.

To reinforce this, some analysts propose that Washington make access to leading-edge American chips conditional on commitments to settle AI-enabled exports in dollars, using dollar-backed stablecoins.

Why computing is not oil

But the comparison with the petrodollar is imperfect. Oil is a relatively homogeneous, universally needed commodity. AI services are diverse, rapidly changing and sometimes replaceable. Open-weight models can be downloaded and operated without recurring payments to an American model provider.

AI-related capital flows are also concentrated in volatile equities and private assets rather than the safe government securities that absorbed much of the petrodollar surpluses. The AI dollar is therefore less a commodity pricing arrangement than a technology and capital market system. Its strength depends heavily on continuing confidence in the profitability of a relatively small number of American companies.

China’s alternative stack

China could disrupt several parts of this system. Open-weight models such as DeepSeek, Qwen and Kimi reduce dependence on dollar-priced American AI services. Chinese-designed processors from Huawei and Cambricon could substitute for Nvidia chips, particularly for less-demanding applications. Alibaba, Baidu and Tencent offer non-American cloud infrastructure, while Chinese policy banks can finance data centres and digital networks in developing countries.

China is also developing alternative payment channels. By June, its Cross-Border Interbank Payment System, or CIPS, had 1,829 direct and indirect financial-institution participants located in 130 countries and regions.

CIPS enables cross-border renminbi clearing and settlement without relying entirely on the traditional dollar-based correspondent-banking system.

The strongest Chinese challenge would combine these different elements: a low-cost Chinese model running on Chinese chips in a Chinese-financed data centre, with cloud charges paid in renminbi through Chinese payment rails. Such a package could appeal to developing countries concerned about cost, US sanctions or excessive dependence on American companies.

Beijing’s constraints

China nevertheless faces serious constraints. Its most advanced chips still lag behind the American frontier, and their production is restricted by limited access to advanced fabrication equipment.

But the gap between leading Chinese and American AI models has narrowed, while Chinese open-weight models compete strongly on cost and accessibility.

Chinese cloud providers still lack the global reach and trust enjoyed by the largest American platforms. More importantly, the renminbi is not fully convertible. Capital controls restrict the movement of money, and China lacks a market for safe, liquid and internationally trusted assets comparable in scale to the Treasury market.

So, although China cannot create an attractive reserve-currency system, it can create a sizeable alternative AI ecosystem that could fragment the AI-dollar network, particularly across the Global South. But replacing it would require major changes to China’s financial system.

America’s threat from within

America’s technological strength is accompanied by growing fiscal weakness. The federal government has run deficits continuously since 2002. Gross federal debt has risen above US$40 trillion and is approximately 125 per cent of gross domestic product, exceeding its post-World War II record.

Concerns about America’s fiscal trajectory are shared by credit rating agencies. S&P downgraded US sovereign debt in 2011. Fitch followed in 2023 and Moody’s in 2025, leaving the US without a top-tier rating from any of the three major agencies.

Without a credible fiscal consolidation plan, persistent deficits and rising interest costs are adding to investor concerns.

Long-term Treasury yields have climbed sharply, with the 30-year yield reaching its highest level since 2007, and the 10-year yield – a key benchmark for mortgages and corporate borrowing costs – has surpassed 5 per cent.

Fiscal worries are not the only cause: inflation, energy prices, monetary policy and strong economic growth also affect yields.

In the face of such risks, investors are demanding greater compensation for holding long-term government debt.

Foreign investors may continue buying American technology shares while reducing their exposure to Treasuries. They can also hedge the currency risk on their equity purchases by selling dollars forward. The US would still receive capital, but the dollar would gain less support because the initial demand for dollars would be partly offset by forward sales.

Meanwhile, central banks may diversify more of their reserves into gold or other currencies. Concerns about Federal Reserve independence, debt-ceiling confrontations and unpredictable policymaking could accelerate that shift.

This would divide the “technology dollar” from the “sovereign dollar”. Investors might remain confident in Nvidia, Google or Microsoft without being equally confident in the dollar’s purchasing power or the long-term management of US public debt.

A more fragile foundation

Investment in AI can temporarily mask fiscal deterioration. Strong equity inflows may help finance the US current-account deficit even as foreign demand for government debt weakens. But this changes the quality of support for the dollar.

Demand for Treasuries is broad, relatively stable and driven by liquidity and safety. Demand for AI equities is concentrated, return-seeking and highly sensitive to valuations. If anticipated AI profits disappoint, foreign capital could retreat just as concerns about US debt are intensifying.

The dollar would then be supported less by confidence in the American state than by confidence in a narrow group of technology companies, which is a more fragile foundation.

Technology cannot substitute for trust

American dominance of chips, cloud computing, AI models, capital markets and dollar stablecoins gives the US a formidable advantage. China can dilute that advantage by building an alternative ecosystem, although its technological and financial limitations constrain the renminbi’s appeal.

The greatest danger to the AI dollar may ultimately come from within the US. Technological leadership can attract capital and postpone the consequences of fiscal indiscipline. It cannot permanently compensate for inflation, the erosion of central bank independence or declining institutional credibility.

AI can give the world another reason to earn and spend dollars. But only sound institutions and credible economic management can persuade it to translate this into enduring confidence in the dollar. An AI dollar alone cannot do that. THE STRAITS TIMES