China looks to dominate the world of AI Tokens and Tokenomics

“Tokens are the new commodity,” said Nvidia CEO Jensen Huang, clad in his iconic leather jacket, at the company’s annual flagship developer conference, GTC, in San Jose, California.

The chip designer’s helmsman wants to recast his company not as a silicon vendor but as the architect of what he calls “Artificial Intelligence (AI) factories”, whose standard product is “token” – and something which increasingly China may be trying to use to de-dollarise the world by transforming plain old AI Tokens into actual units of currency.

While Nvidia is busy writing the rules of a new token economy, a parallel debate is emerging in China around the idea of “token exports.” AI-generated intelligence is essentially a tradeable good measured by token, and China is positioning itself across the value chain – from energy and computing power to models and output.

In this article we examine how “Tokenomics” works – and why some believe China has the edge. In AI, a token is the smallest unit of data an AI model processes or generates. Unlike crypto tokens – which encode ownership or speculative value – AI tokens are purely computational: they are what users pay for, and what models produce.

If Satoshi Nakamoto, the presumed creator of Bitcoin, gave the world a token you could trade, Huang is selling one that works.

The Future and Deep Future of AI 2040, a keynote by futurist Matthew Griffin

As token consumption surges – driven by multimodal models and the rise of AI agents and all kinds of applications – the production efficiency and price of tokens are becoming increasingly critical. Huang argued that tokens were to the AI era what barrels were to oil: a commodity to be produced, priced and competed over. Data centres become factories, revenue is measured in tokens per watt, and companies may soon manage token budgets much as they do headcount.

China’s tech groups are moving in a similar direction. Alibaba, which is repositioning itself as an AI-driven company, last week reorganised its AI operations into a new top-level unit, the “Alibaba Token Hub”, with the mission to create, deliver and apply tokens.

Tokens are a resource, and like every resource, they have a supply constraint. Producing them at scale requires chips, cooling and essentially electricity.

That reality has triggered an infrastructure boom of historic proportions. US tech giants are committing hundreds of billions of dollars to build data centres, while in some regions grid connection queues already stretch to 2028.

Tokens are consumed via models. Weekly data from OpenRouter showed that the four most-used models by token consumption are all Chinese – from StepFun, MiniMax, Xiaomi and DeepSeek – with Zhipu also in the top 10. Together, the top 10 account for 18.8 trillion tokens. There are also popular Chinese models, like Alibaba’s Qwen series, that are not in the chart because of their open source nature.

The reason is largely cost. API pricing for models such as Zhipu’s GLM-5 can be an order of magnitude lower than premium Western models like Claude 4.6 Opus. As token usage scales, price becomes a decisive factor.

China still lags the US at the cutting edge of chips. But in a token economy – where the key metrics are output per watt and cost per million tokens – it is building a different kind of advantage.

China is the world’s largest electricity producer, and its power capacity has expanded at a pace that has repeatedly outstripped official targets.

By the end of 2025, the country’s total installed generating capacity was 3.89 billion kilowatts – nearly three times that of the US, according to their respective energy agencies. Solar alone grew 35 per cent last year, while wind increased 23 per cent.

China’s government work report this year introduced “compute-electricity synergy” as a national priority, aiming to align data centre deployment with energy resources and address geographic mismatches between supply and demand – through projects such as west-to-east power transmission and the “East Data, West Compute” programme.

What makes this advantage harder to replicate is the depth of China’s supply chain. The country dominated the manufacturing of key grid and energy components – including transformers, inverters, switching equipment and solar panels – many of which were now being sourced by companies such as Google, Tesla and Amazon, reports said.

First published on Fanatical Futurist on 16 March 2026.

311 Intelligence

What this means

Analysis by the 311 Strategic Intelligence Unit

The big shift is from AI as software to AI as a commodity measured in tokens. Jensen Huang argues tokens are to the AI era what barrels were to oil, and Alibaba has built a whole unit around them. If output is priced per token, the winners are those with the cheapest power and the cheapest models, not only the best chips. We rate AI Inference Cost Collapse, the price of running AI at a given quality falling roughly tenfold or more each year for most tasks, an Act now trend.

What it means by sector.

• Technology and cloud: the four most-used models on OpenRouter are Chinese, and API pricing for models such as Zhipu's GLM-5 can be an order of magnitude lower than premium Western models. We track Chinese AI in Global South, cheap, open Chinese models gaining fast across the developing world, as a Prepare trend with wide use by 2029.
• Energy and utilities: producing tokens at scale needs chips, cooling and electricity. China's generating capacity is nearly three times that of the US, and we track Electrostate Economies, whole economies shifting to home-made clean electricity, as an Explore trend with wide use by 2035.
• Every organisation using AI: token consumption is surging with multimodal models and AI agents. We expect Agentic Artificial Intelligence, AI that can decide and act on its own to reach a goal, to reach wide use by 2028. Large Language Models already reached wide use in 2024.

What to do now.

• Measure your AI use in tokens and cost per million tokens, and set token budgets alongside headcount budgets.
• Compare open and low-cost models against premium ones for each task, since price becomes decisive as usage scales.
• Factor energy supply and grid access into any data centre or AI capacity plan.

The Starburst technologies behind this story

From the 311 Technology Starburst (first product, expected wide use and readiness):

Technology	Category	First product	Wide use	Action	TRL
Large Language Models	Intelligence	2020	2024	Act now	9
Agentic Artificial Intelligence	Intelligence	2024	2028	Prepare	8
Photovoltaics	Energy	1954	2026	Act now	9
Grid Forming Inverters	Energy	2021	2031	Explore	9

And the trends they feed, from the 311 Trends Starburst:

Trend	Wide use	Action
AI Inference Cost Collapse	2026	Act now
Chinese AI in Global South	2029	Prepare
Electrostate Economies	2035	Explore
Global Reserve Currency Wars	2035	Explore

How this moves the trend

China has been building toward this for years. In 2024 it planned a 50% boost in national computing power, and in 2025 DeepSeek's low-cost model tanked US stocks by over $1 trillion. The same year Google's AI usage smashed past a quadrillion tokens in a month, and in 2026 Jensen Huang floated paying Nvidia engineers half their annual salary in AI tokens. Tokens are fast becoming the unit the AI economy is measured in.

Go deeper with 311

• The 311 Starburst Centre: every technology and trend above, with timelines and readiness.
• 311 Vantage: ask what the AI token economy means for your organisation and get a plain English executive brief.
• The Exponential Technology Codex and the Annual Trends Codex: the technologies and trends reshaping every industry.
• Book a briefing or keynote on the future of AI and the token economy for your board or leadership team.