The token that I saw because I believed in it in the past can now be seen without believing. It's the next one after watts, amps, bits.

In January 2009, an anonymous person invented something called "token". You invest computing power and obtain tokens, which are circulated, priced, and traded in a consensus network. The entire crypto economy was born out of this. More than ten years have passed, and people are still debating whether this token has any value.

In March 2025, a man in leather clothes redefined another thing called token. You invest computing power and produce tokens, which are immediately consumed in a process of AI inference & reasoning: thinking, reasoning, writing code, and making decisions. The entire AI economy accelerates as a result. No one is debating whether this token has value because you just used millions of them this morning.

Two tokens, the same name, the same underlying structure: computing power goes in, and valuable things come out.

In March 2026, I sat in the NVIDIA GTC venue and listened to a keynote speech by Jen-Hsun Huang that had almost no product. Yes, he released Vera Rubin, a combination CPU and GPU. But this time, he did not talk about chip parameters or process technology. He talked about a complete set of economics about token production, pricing and consumption——

Which model corresponds to which token speed; which token speed corresponds to which pricing range; which pricing range requires what level of hardware to support it.

He even helped the CEOs in the audience and the decision-makers who controlled the company's checkbook prepare a data center computing power allocation plan: 25% for the free tier, 25% for the mid-range, 25% for the high-end, and 25% for the high-premium tier.

Yes, this time he did not specifically sell which GPU set, just like he did with Blackwell two years ago. But this time, he's selling something bigger. After two hours, I think what he wanted to say most was actually: Welcome to consume tokens, and only Nvidia's factory could produce.

At this moment, I realized that this man, and the man who dug up the first token 17 years ago by Anonymous, were doing the same thing in structure.

Same set of conversion rules

The anonymous person with the pseudonym "Satoshi Nakamoto" wrote a nine-page white paper in 2008 and designed a set of rules: invest computing power, complete a mathematical proof (Proof of Work), and receive crypto tokens as a reward.

The subtlety of this rule is that it does not require anyone to trust anyone - as long as you accept this set of rules, you will automatically become a participant in this economy. This rule is right, after all, it brings so many cunning people together.

On the stage of GTC 2026, Jen-Hsun Huang did something structurally identical.

He showed a graph that highlights the relationship and tension between inference efficiency and token consumption: the Y-axis is throughput (how many tokens are produced per megawatt of power consumption), and the X-axis is interactivity (the speed of tokens perceived by each user). Then, he marked five pricing brackets under the

This picture can almost be used as the cover of Huang Renxun's "Token Economics" white paper.

Satoshi Nakamoto defined "what is valuable computation" - completing SHA-256 hash collisions is valuable. Huang Renxun defined "what is valuable reasoning" - under given power consumption constraints, it is valuable to produce tokens for specific scenarios at a specific speed.

Neither Satoshi Nakamoto nor Huang Jen-Hsun directly produces tokens. What they define are token production rules and pricing mechanisms.

What Huang said on the stage can almost be written directly into the summary of the token economics white paper——

Tokens are the new commodity, and like all commodities, once it reaches an inflection, once it becomes mature, it will segment into different parts.

Token is the new commodity. Commodities will naturally stratify as they mature. He is not describing the current situation, he is predicting a market structure, and then accurately laying out his own hardware product line on every layer of this structure.

The production processes of the two tokens even have a semantic symmetry: mining is called mining, and reasoning is called inference.

The essence of mining and reasoning is to turn electricity into money. Miners spend electricity fees to mine crypto tokens and then sell them. Inference models and AI Agents spend electricity fees to generate AI tokens and then sell them to developers at millions of dollars. The middle link is different, but both ends are the same: the meter on the left and the income on the right.

Two ways to write scarcity

The most important design decision Satoshi Nakamoto ever made was not Proof of Work, but the upper limit of 21 million Bitcoins. He used code to create artificial scarcity—no matter how many miners flood in, the total number of Bitcoins will never exceed 21 million. This scarcity is the value anchor of the entire crypto economy.

Huang Renxun used the laws of physics to create natural scarcity. He said——

"You still have to build a gigawatt data center. You still have to build a gigawatt factory, and that one gigawatt factory for 15 years amortized... is about $40 even billion when you put nothing on it. It's $40 billion. You better make for darn sure you put the best computer system on that thing so that you can have the best token cost."

A 1GW data center will never become a 2GW. This is not a code limitation, this is a law of physics.

Land, electricity, heat dissipation - each has a physical upper limit. How many tokens you can produce in the 15-year life cycle of this factory that you spent 40 billion US dollars to build depends entirely on what computing architecture you put in it.

Satoshi's scarcity can be forked. If you don’t like the upper limit of 21 million coins, fork a new chain, change it to 200 million coins, call it Ethereum or something, whatever you want, and post a white paper by the way. And people do it, and enjoy it.

The scarcity created by Lao Huang cannot be forked. After all, you can’t fork the second law of thermodynamics, you can’t fork the grid capacity of a city, you can’t fork the physical area of ​​a piece of land.

But whether it is Satoshi Nakamoto or Huang Jen-Hsun, the scarcity they created has led to the same result: a hardware arms race.

The history of mining is: CPU→GPU→FPGA→ASIC. Each generation of specialized hardware renders the previous generation scrap. And the history of AI training and inference is repeating itself: Hopper→Blackwell→Vera Rubin→Groq LPU. It starts with general-purpose hardware and ends with special-purpose hardware. The Groq LPU that Lao Huang showed at GTC this year is a deterministic data flow processor released after acquiring Groq. Static compilation, compiler scheduling, no dynamic scheduling, 500MB on-chip SRAM - its architectural philosophy is the ASIC in the field of reasoning. Just do one thing, but do it to the best of your ability.

Interestingly: GPUs played a key role in both waves.

Around 2013, miners discovered that GPUs were more suitable for mining crypto tokens than CPUs, and Nvidia graphics cards were sold out. Ten years later, researchers discovered that GPUs were the best tool for training and inferring AI models, and Nvidia data center cards were once again sold out. As a processor category, GPU has served two generations of token economy.

The difference is that Nvidia benefited passively the first time, and then nothing happened. The second time, when the main battlefield of AI computing power consumption switched from pre-training to inference testing, NVIDIA quickly seized the opportunity to proactively design the entire game and became the writer of AI game rules.

The most profitable shovel in the world

The most profitable person in the gold rush was not the gold digger, but Levi Strauss, the shovel seller. The most profitable people in the mining boom are not the miners, but Bitmain and Wu Jihan who sell mining machines. The most profitable thing in the AI ​​pre-training and inference wave is not the base model and Agent, but NVIDIA, which sells GPUs.

But to be honest, the roles played by Bitmain and NVIDIA in their respective industries are no longer the same as those in Japan.

Bitmain only sells mining machines, and Nvidia was once a supplier to Bitmain. When you buy a mining machine, what coins you mine, which mining pool you go to, and at what price you sell them have nothing to do with Bitmain. It is a pure hardware supplier and makes one-time equipment profits.

NVIDIA is different. He not only sells hardware, but now, especially since the explosion of inference-side AI in 2025, it deeply defines what should be mined with this GPU, how to price tokens, who to sell tokens to, and how data centers should allocate computing power... These are all in Huang's speech PPT: He divided the market into five tiers, each tier corresponds to what model, context length, interaction speed and price... NVIDIA has standardized and formatted the future market where AI inference drives everything.

Around 2018, global computing power was concentrated in several large mining pools—F2Pool, Antpool, and BTC.com—which competed with each other for computing power shares, but the source of mining machines was highly concentrated in Bitmain.

Like today's Nvidia, 60% of revenue comes from competing "hyperscalers" such as AWS, Azure, GCP, Oracle, CoreWeave, and 40% comes from decentralized AI Natives, sovereign AI projects, and enterprise customers. Large “mining pools” contribute the majority of revenue, while small “miners” provide resilience and diversification.

The structures of the two ecologies are exactly the same. But Bitmain later encountered competitors - Shenma Mining Machinery, Innosilicon Technology, and Canaan Technology, all of which were encroaching on its share. Miners are relatively simple ASIC designs, and there are opportunities for chasers. It seems to be getting harder and harder to shake NVIDIA: 20 years of CUDA ecosystem, hundreds of millions of GPU installed base, NVLink six-generation interconnect technology, Groq integrated decoupled inference architecture - NVIDIA's technical complexity and ecological barriers make most competitive tools ineffective.

This may have to last 20 years.

The fundamental bifurcation of the two tokens

What makes cryptocurrency and AI training and reasoning tokens fundamentally different is the motivation and psychology of people using them.

The demand side of Crypto tokens is speculation. No one “needs” Bitcoin to get work done. All white papers claiming that blockchain tokens can help you solve your problems are liars. You hold crypto because you believe someone will buy it from you at a higher price in the future. Bitcoin's value comes from a self-fulfilling prophecy: enough people believe it has value, and it has value. This is an economy of faith.

The demand side of AI token is productivity. Nestlé needs tokens to make supply chain decisions - its supply chain data is refreshed from 15 minutes to 3 minutes, reducing costs by 83%. This value can be directly mapped to P&L. 100% of Nvidia's engineers already need tokens to write code instead of rubbing their hands; the research team needs tokens to engage in scientific research. You don’t need to believe that the token is valuable, you just need to use it, and the value will prove itself during use.

This is the most essential difference between the two tokens. Crypto tokens are produced to be held and traded - their value lies in not being used. AI tokens are produced to be consumed immediately—its value lies in the moment it is used.

One is digital gold, which becomes more valuable the more you hoard it; the other is digital electricity, which is burned as soon as it is produced.

This difference determines that the AI token economy will not bubble like the crypto token economy. Bitcoin rises and falls as the price of the speculative commodity is driven by sentiment. But the price of tokens is driven by usage and production costs. As long as AI continues to be useful—as long as people are still using Claude Code to write code, ChatGPT to write reports, and Agent to run business processes, the demand for tokens will not collapse. It does not depend on faith, but on inseparability.

In 2008, the Bitcoin white paper needed to reiterate why a decentralized electronic cash system was valuable. 17 years have passed and people are still fighting.

In 2026, token economics has not triggered any controversy, and it has become a consensus without even needing argumentation. When Lao Huang stood on the GTC stage and said "tokens are the new commodity", no one questioned it. Because everyone sitting in the audience has spent millions of tokens using Claude Code or ChatGPT this morning. They don’t need to be convinced that the token has value – their credit card statement has already proven it.

In this sense, Lao Huang is really a copy of Satoshi Nakamoto, the copy who left behind Satoshi to monopolize the production of mining machines, defined the usage scenarios and usage specifications of tokens, and held an annual show at the SAP Center in San Jose to tell people how powerful the next generation of "mining machines" that support AI training and reasoning is.

Satoshi Nakamoto has the charm of desire and prudence. He designs the rules, hands them over to the code, and then disappears. This is the romance of cypherpunk. Huang is more like a businessman than any other scientist. He designed the rules, maintained them personally, and constantly added bricks and mortar to build his own moat.

The token you saw in the past because you believed it can now be seen without believing. It's the next one after watts, amps, bits.