Tokens are expensive, and the cost is painful.
This is not only the sentiment of those currently obsessed with Vibe Coding, but even Silicon Valley giants who previously fervently advocated for Tokenmaxxing have started imposing token limits on their own employees.
However, a counterintuitive point is that students currently using AI subscriptions are actually using tokens that have already been subsidized by the major AI companies. The highest subsidy might even be a staggering 70 times the subscription fee!
What's more concerning is that both OpenAI and Anthropic, the two leading AI companies, have entered the final sprint towards their IPOs. After these companies go public,
will they follow the pattern of the internet era's "subsidy wars," where the remaining companies begin raising customer prices, causing token prices to return to rationality?
The good news is, this might not happen. Recently, Bill Maris, the founder of Google Ventures, raised a question on the All-in podcast:
If Google decided to slash token prices by another 80%, how would OpenAI and Anthropic respond?
Coincidentally, not long ago, the startup team Agnes AI, in a livestream with GeekPark, explained in detail the possible advent of the "Token Free Era."
So, will the price of tokens increase or decrease in the future? And what does this mean for those already addicted to AI?
01 The Token Subsidy War is Fierce
Why is it said that current token prices are not actually expensive?
Because, at least in terms of AI subscriptions, the current prices offered by various AI companies are already "rock-bottom prices" after subsidies.
Recently, SemiAnalysis conducted a detailed evaluation comparing the actual token consumption value against subscription fees in the subscription models of OpenAI and Anthropic.
SemiAnalysis did something simple yet effective—actually using AI to complete various tasks under different subscription plans on various AI platforms, then using the public API pricing to calculate back how much those tasks' tokens were worth. The results are as follows:
Note a pattern: the more expensive the plan, the higher the subsidy multiple. This itself indicates that these high-end plans are not meant for profit—they are a form of "reverse pricing," using the most aggressive losses to retain the heaviest users. Because heavy users are developers, enterprise decision-makers. Once they are locked into a platform, they bring along entire teams and product lines.
With losses this severe, why continue? The standard answer is: burn money first to gain scale, then raise prices to recoup after achieving scale. Mobile internet played this game—DiDi and Uber subsidized hundreds of billions of RMB in ride fares, and fares increased after subsidies ended; Meituan subsidized countless food deliveries, and delivery fees increased after subsidies ended. This logic hinges on one key premise: a lock-in effect was established during the subsidy period.
DiDi could raise prices because drivers couldn't leave the order flow on the platform, and passengers couldn't leave the drivers on the platform. Meituan could raise prices because merchants couldn't leave its traffic and delivery network. When subsidies ended, users were already "locked" into the ecosystem, making switching costs extremely high.
But the AI war differs fundamentally from the internet—Tokens have almost no lock-in effect.
If Claude raises prices, developers can migrate API calls to GPT or Gemini within a day—interfaces from different companies are becoming increasingly standardized, with many development frameworks even having built-in multi-model switching functions. It's even simpler for regular users: just change a URL. AI isn't like ride-hailing with its local driver network, or food delivery with its logistics system, or social media with its friend networks. A token is a token; no matter who produces it, it's the same thing.
This means that once subsidies stop, users can be lost instantly. Subsidies aren't "building moats"; they're more like "maintaining a heartbeat"—as soon as someone offers a lower price, users run.
And this hasn't even accounted for a new variable that's making everyone's bills spiral out of control: AI Agents.
When you chat with ChatGPT, a single conversation might consume a few thousand tokens. But when you have an AI Agent perform a complex task—writing a piece of code and debugging it automatically, analyzing a dozens-page document and generating a report—a single round can consume 5 to 30 times the tokens of a normal conversation. Developers have tested that on a $100 Claude Max plan, a single Agent programming session can burn through nearly $100 worth of tokens. Uber's CTO recently revealed that the company burned through its entire 2026 AI budget in just four months.
The question is, can such a Token Subsidy War continue? Who might be the ones left standing after the chaotic battle?
Bill Maris believes the answer is obviously the traditional giants.
02 Token as a Weapon
To understand the true brutality of this subsidy war, one must first see a structural asymmetry—the sources of ammunition for each combatant are completely different.
Google's annual ad revenue exceeds $300 billion. This isn't money from investors, not money raised and burned, but an automatic money-printing machine running daily. Billions of people worldwide open its search engine, watch YouTube, and use Gmail every day, and ad revenue flows into its account automatically. It doesn't need roadshows, doesn't need to please analysts, doesn't need to explain to anyone why it's spending this money.
Google using ad profits to subsidize AI tokens is like someone with an oil well fighting a gas station price war—his oil comes from his own land, while his opponents' oil is bought with bank loans.
OpenAI and Anthropic are those buying oil with loans.
OpenAI has raised over $180 billion cumulatively, with its latest valuation exceeding $850 billion. Anthropic has raised over $130 billion. This money comes from venture capital and strategic investors—they aren't giving money as charity; they expect these companies to go public, expect to get rich returns upon exit.
And after going public, the real trouble begins. Going public means financial statements are public to the world. Every quarter, Wall Street analysts will scrutinize revenue, profit, customer acquisition cost, and marginal cost. When they calculate that you actually lose $70 for every $1 of subscription fee you receive—even the most glorious growth story won't support the stock price.
Bill Maris spelled out this logic bluntly on the podcast. His exact words were: "If I were Google, and decided to arbitrarily cut token prices by 80%, what would happen to OpenAI and Anthropic's business models?"
The host pressed, asking how likely that was. Maris didn't hesitate: "100%. Capital as a weapon, tokens as a weapon."
This isn't analyst speculation. Bill Maris is the founder and CEO of Google Ventures, also a Google VP for Special Projects, having incubated Waymo and Google X. Everyone present understood: this isn't a hypothesis; this is him having seen firsthand how Google fights wars.
He painted a simple scenario: Google announces an 80% price cut for the Gemini API. What would enterprise customers do? If the product quality is similar—in many benchmark tests Gemini is already on par with Claude and GPT—but the price is four-fifths cheaper, would you continue using the expensive one?
Maris gave his own answer: "If you're a company, and you can pay 80% less at Google and Gemini for basically the same product, why wouldn't you? And then the pressure on those companies becomes very severe."
And OpenAI and Anthropic have almost no symmetrical means of retaliation. They cannot follow suit with price cuts—they have no money-printing machine; every dollar is investor money. They also cannot maintain a premium based on a technical gap—the gap between large models is shrinking rapidly; you might lead by three months today, and be caught up in three months. This isn't like the generational technological gap between the iPhone and Nokia. The moat between AI models is more like a sand dike; the tide rises and washes over it.
In Bill's narrative, Google has a big advantage, but in the AI world, can Google really monopolize? Meta can open-source a free model at any time, China has DeepSeek and ByteDance, Amazon is pushing its own model. When you drive token prices down to cabbage prices, the competitors don't disappear—they are also cutting prices.
The AI war might have no winner.
03 The "Infinite Game" of Tokens?
Even those unfamiliar with history might make the following judgment about the ultimate outcome of the current AI war:
The first is the "Internet Service" script—the DiDi story, the Amazon story: subsidize first, then monopolize, then raise prices to harvest. In this script, today's price war is just the prologue; one or two winners will eventually dominate most of the market and gain pricing power. If this is the case, the huge losses today are a worthwhile investment—just like Amazon lost money for two decades before finally becoming the dual titan of e-commerce and cloud computing.
The second is the "Utilities" script. Tokens become a standardized basic resource, like electricity, bandwidth, cloud storage. No one can maintain pricing power long-term because product differentiation is too small, switching costs too low. Competition drives prices infinitely towards the cost line, with profit margins approaching zero. Ultimately, governments might step in for regulation—just as they did with electricity and telecom a hundred years ago.
The dividing line between the two scripts depends on one word:
Lock-in.
DiDi could raise prices because passengers are locked into the driver network, and drivers are locked into the order flow. Amazon could raise prices because merchants are locked into its logistics and traffic ecosystem.
The lock-in effect is the cornerstone of the "lose first, earn later" model.
But AI tokens—as argued repeatedly above—have almost no lock-in. APIs are standardized, switching costs are practically zero. The core condition for the first script doesn't exist for the product that is tokens.
If the second script, the utilities/infrastructure endgame, is closer to reality, then what we are witnessing is not a war that will eventually have a winner, but an attrition race with no end.
Meituan founder Wang Xing once described this competitive state. His insight was: some competitions have no concept of "winning." The goal of participants isn't to defeat opponents, but to ensure they remain at the table. Because as long as you are still at the table, you can continue raising funds, hiring, iterating. Leaving the table is the only way to lose.
Re-examining today's AI landscape with this framework, many seemingly contradictory things suddenly become clear.
OpenAI's latest valuation exceeding $800 billion isn't because training models requires that much money. It needs that much money to continue fighting the price war. Raising funds isn't to win; it's to "qualify to keep fighting."
Google preparing to cut token prices by 80% isn't to eliminate OpenAI and Anthropic. It's to ensure it remains a core player in the AI era—just as it once ensured it wouldn't be left off the table in the mobile era by giving away Android for free.
And Anthropic raising the API pricing of its latest flagship model Fable 5 to double that of its predecessor—$10 per million tokens for input, $50 per million for output—seemingly "raising prices," is actually proactively filtering for enterprise customers willing to pay for high-end capabilities, because deep down it knows: the consumer-side subsidy war cannot be won against Google.
Every round of price war expands the scale of AI usage. Expanding scale means more data, more scenarios, more developers flooding into the ecosystem. This, in turn, makes all participants' models stronger. Combatants use the war itself to attract resources to upgrade themselves—this isn't a zero-sum, life-and-death battle, but a process where everyone becomes stronger through competition, but also one where no one is likely to make huge profits.
Doesn't this sound like what the electricity industry eventually became?
140 years ago, Edison and Westinghouse both thought they were fighting for a winner-takes-all market. They staked their entire fortunes, betting that "whoever defines the standard for electricity owns electricity." But the fate of electricity teaches us a simple truth:
When a technology is important enough, general enough, standardized enough, it no longer belongs to any single company. It belongs to the infrastructure.
On the surface, the AI competition seems to be Google vs. OpenAI vs. Anthropic, a contest of model capabilities, a comparison of fundraising scale. But zooming out, the real role of this competition is: it is accelerating the push of AI towards a level of infrastructure that no single company can monopolize.
When Bill Maris says "100% will happen," he might not just be predicting that Google will cut prices. He might be unconsciously predicting a larger trend—in the AI world, tokens ultimately won't belong to anyone. Just like no one "owns" electricity today.
For OpenAI and Anthropic, this means something unsettling: even with technological leadership, even after raising astronomical sums, the future of "making big money from AI" they are chasing might never have existed. They aren't facing a temporary price war, but a structural fate—the thing they are striving to build is, in essence, likely the next generation's water, electricity, and highways.
For users, to some extent, this might be good news. Because as long as the Token Subsidy War continues, people can still enjoy the "good deal" of $20 cost for $400 worth of computing power.
This article is from WeChat Official Account "GeekPark" (ID: geekpark), Author: Yu Hangyuan (Astronaut Ape)
