appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Jeffrey Wu

A quiet but consequential shift is reshaping the global artificial intelligence competition, and it has little to do with which country builds the most powerful model.
Jensen Huang did not mean to describe a geopolitical strategy. But when Nvidia’s chief executive declared, “Your workload is inference, your tokens are your commodity, and that compute is your revenue,” he was articulating, from the supply side, something China had concluded from the other direction.
To understand why, start with a basic concept. Tokens are the fundamental units AI models use to process and generate language: every word, response and automated task breaks down into them. Cloud providers charge by the token the way utilities charge by the kilowatt-hour. Whoever produces them most cheaply, at the greatest scale, holds an advantage in the AI economy like how cheap steel once decided industrial supremacy.
On OpenRouter, a widely used platform for developer AI access, Chinese models have surpassed American ones in token consumption for the first time: 5.16 trillion tokens against 2.7 trillion in a single week. By mid-March, Chinese models accounted for 36 per cent of OpenRouter’s global volume, with a weekly consumption of 7.36 trillion tokens.
Washington has spent three years building the most comprehensive technology export control regime since the Cold War, designed for a world where competitive advantage travels in hardware: chips that can be counted, shipments that can be blocked, supply chains that can be pressured. A different contest is now taking shape alongside it, one for which the existing toolkit has no architecture.
Industrial eras have been defined not only by their dominant technologies, but by the units used to measure them. Tonnes of steel told you who was industrialising. Barrels of oil told you who held leverage. Gross domestic product, constructed in the 1930s, defined what economic activity meant, and governments built policy architectures around maximising it.


Last month, at the China Development Forum in Beijing, Liu Liehong, head of China’s National Data Administration, gave tokens an official name, ciyuan, and declared them to be value anchors of the intelligent era and “settlement units”. Daily consumption, he reported, had reached 140 trillion tokens – just two years earlier, the figure was 100 billion.
No other government has elevated an AI engineering concept into a national economic indicator. China just proposed a new unit of account for the intelligent era.
And it is being assembled in physical infrastructure. Alibaba Group Holding (which owns the South China Morning Post) has launched a 10,000-card computing cluster powered by domestically developed Zhenwu chips in Shenzhen. Huawei Technologies’ Ascend 910C cluster, also in Shenzhen, delivers 11,000 petaflops of capacity. These are not demonstrations of chip capability alone.
They are statements about productive scale, capacity built ahead of the demand the system intends to create – closer in logic to the American railroad buildout of the 1870s than to a nation scrambling to catch up.
The cost advantage in producing tokens is structural. China has designated “compute-electricity synergy” a national priority, explicitly linking cheap renewable energy to AI competitiveness. Chinese models charge roughly US$2-3 per million output tokens against about US$15 for leading Western equivalents.


That gap is amplified by AI agent applications – software that executes complex tasks autonomously – which consume tokens at 10 to 100 times the rate of ordinary conversation. When per-session costs surged, developers migrated towards price. The migration was not ideological.
Japan’s very large scale integration (VLSI) semiconductor project offers the most instructive historical parallel, and the sharpest limit. In the late 1970s, the state-industry effort achieved cost-competitive production and aggressive exports. Japan’s dynamic random-access memory (DRAM) chips captured the global market: the US share collapsed from 70 per cent to 20 per cent between 1978 and 1986.
The response was the 1986 US-Japan Semiconductor Agreement, which imposed a price floor and market share monitoring. Those mechanisms worked because the export was visible: chips counted at borders, priced at customs, subjected to treaty enforcement. Token export has no equivalent exposure. You cannot impound an application programming interface (API) call. There is no minimum pricing mechanism for inference delivered through undersea fibre.
That advantage has a ceiling. Enterprise workloads involving sensitive data carry regulatory and reputational costs that price alone cannot overcome. Whether it scales beyond developers into enterprise adoption across Southeast Asia and the Global South determines whether this is structural or an early adopter phenomenon.


The domestic numbers warrant scrutiny too. ByteDance’s Volcano Engine alone reports 120 trillion daily tokens. JPMorgan projects a 370-fold growth in Chinese inference consumption by 2030. Yet token volume does not equal economic value. Much of it is subsidy-driven. The Soviet Union tracked steel output with comparable enthusiasm; optimising for the metric eventually distorted the economy.
The unit of account is rarely set by the most innovative economy. It is set by the economy with the most users, deepest infrastructure and lowest adoption cost. The Bretton Woods system was not an act of cleverness, but a recognition of productive weight. China is building the preconditions for a parallel claim: domestic infrastructure at scale, cost-efficient inference, an export mechanism invisible to trade frameworks and a national accounting system with the token at its centre.
The question is no longer only who builds the most powerful AI. It is who defines the unit by which AI output is measured – and who is using that unit when the standard gets set.


As AI safety fears mount, can a US-China hotline prevent a global crisis?

In a China first, low-altitude sector hits turbulence as subsidies face reality check

Expect ‘high tide’ of Chinese biotech deals as threat of US curbs fades: analysts

Asia investors demand AI revenue proof as focus shifts from exposure to earnings: BofA

China technology

US controls chips in the AI race, but China controls the scoreboard

A lobster-shaped pendant at an OpenClaw event in Beijing on March 22. At a time when AI agent platforms like OpenClaw are taking China by storm, the Chinese government has proposed a new unit of account for the intelligent era. Photo: Xinhua


What is ‘tokenomics’ and how would China gain the edge in artificial intelligence era?

Jensen Huang at Nvidia’s annual GTC conference in San Jose, California, on March 17. Photo: AFP

AI-generated intelligence is essentially a tradeable good, and China is positioning itself across the value chain – from energy and computing power to models and output. Photo: Shutterstock

China’s power capacity has expanded at a pace that has repeatedly outstripped official targets. Photo: Xinhua

China names AI tokens after the yuan. Should the US worry for the dollar?

China has designated ciyuan as the translation for “token”, during an official’s speech on AI development that linked tech supply with commercial demand. Photo: Reuters

How prolonged Iran war could disrupt Asia tech industry from chipmaking to AI data centres

An LNG tanker is moored in Kawasaki, Kanagawa prefecture, Japan, on April 8, 2026. Photo: AFP

The Ras Laffan complex in Qatar before the attacks in March. Photo: AFP

Opinion

China Opinion

World Opinion

News

China

Science

China Future Tech

Science & Technology

Economy, Trade & Business

Tech

Policy

Tech War

Whoever produces tokens cheaply, at scale, has an advantage in the AI economy, just as cheap steel once decided industrial supremacy


Whoever produces tokens cheaply, at scale, has an advantage in the AI economy, just as cheap steel once decided industrial supremacy.


Artificial intelligence (AI)

AI: Industry Trends

Technology

AI: Companies

Nvidia

Semiconductors

US-China tech war

Infrastructure

AI: Geopolitics

AI: Policy

United States

Jensen Huang

Expresses the views, ideas and conclusions of the author. (Eg. Enough of Crazy Rich Asians, think of crazy poor Hongkongers.)


US controls chips in the AI race, but China controls the scoreboard

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Whoever produces tokens cheaply, at scale, has an advantage in the AI economy, just as cheap steel once decided industrial supremacy

Whoever produces tokens cheaply, at scale, has an advantage in the AI economy, just as cheap steel once decided industrial supremacy