Is the "Token Subsidy War" among AI giants coming to an end?

Tokens are expensive; burning them hurts.

This is not only the voice of those currently obsessed with Vibe Coding, but even the big Silicon Valley companies that previously promoted Tokenmaxxing are now starting to impose Token restrictions on their employees.

But an counterintuitive point is that, for those using AI subscriptions now, the tokens you’re using have actually been subsidized by AI giants, with the highest subsidies possibly reaching 70 times the subscription fee!

What’s even more worrying is that both OpenAI and Anthropic, the leading AI companies, are already in the IPO sprint phase. Once these two companies go public,

will they, like during the “subsidy wars” of the internet era, see the remaining companies raise their prices, bringing Token prices back to rational levels?

The good news is, this scenario might not happen. Recently, Bill Maris, founder of Google Ventures, posed a question on the All-in podcast:

If Google decides to cut token prices by another 80%, how will OpenAI and Anthropic respond?

Coincidentally, not long ago, startup Agnes AI explained in a live session with GeekPark what the coming “Free Token Era” might look like.

So, will the future Token prices go up or down? And what does this mean for people already addicted to AI?

Token subsidies are already smoking

Why do I say that Token prices are actually not expensive right now?

Because, at least in AI subscription models, the current prices of various AI companies are already “rock-bottom” after subsidies.

Recently, SemiAnalysis conducted a detailed comparison of the actual token consumption value versus subscription fees under OpenAI and Anthropic’s subscription plans.

SemiAnalysis did a simple but effective thing—using AI to complete various tasks under different platform plans, then calculating the token value based on API published prices. The results are as follows:

Notice a pattern: the more expensive the plan, the higher the subsidy multiple. This alone indicates that these high-end plans are not meant for profit—they are a form of “reverse pricing,” using aggressive losses to retain heavy users. Because heavy users are developers and enterprise decision-makers, once they are tied to a platform, they can bring along entire teams and product lines.

Why do they do this at such a loss? The standard answer is: burn money first to scale up, then raise prices to recover. This is how mobile internet works—Didi and Uber subsidized hundreds of billions of RMB in ride costs; after subsidies ended, fares increased; Meituan subsidized countless meals, and after subsidies ended, delivery fees rose. The key premise for this logic is: subsidies create lock-in effects.

Didi can raise prices because drivers are locked into the platform’s order flow, and passengers are locked into the driver network. Meituan can raise prices because merchants rely on its traffic and delivery network. After subsidies end, users are “locked” into the ecosystem, with high switching costs.

But in the AI war, there’s a fundamental difference from the internet—Token almost has no lock-in effect.

If Claude raises prices, developers can migrate API calls to GPT or Gemini in a day—interfaces are becoming increasingly standardized, and many frameworks even support multi-model switching. For ordinary users, it’s even simpler: just change the URL. AI doesn’t have a local driver network like ride-hailing, a delivery system like food delivery, or a friend network like social media. Tokens are tokens—regardless of who produces them, they are the same thing.

This means that once subsidies stop, users can instantly lose their engagement. Subsidies are not about “building barriers,” but more like “maintaining heartbeat”—as long as someone offers a lower price, users will leave.

And that’s not counting a new variable that’s causing everyone’s bills to spiral out of control: AI Agents.

When you chat with ChatGPT, a single conversation might consume a few thousand tokens. But when you ask an AI Agent to perform a complex task—write code and debug automatically, analyze a 30-page document and generate a report—the token consumption can be 5 to 30 times that of a normal conversation. Some developers have tested that on a $100 Claude Max plan, a single Agent programming session can burn 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 this Token subsidy war continue? Who might still be standing after the chaos?

Bill Maris believes the answer is clearly the traditional giants.

Token as a weapon

To understand the brutal reality of this subsidy war, we first need to see a structural asymmetry—the sources of ammunition for each side are completely different.

Google’s annual ad revenue exceeds $300 billion. This isn’t money from investors or fundraising, but an automated printing press running every day. Billions of people worldwide open search engines, watch YouTube, use Gmail daily, and ad fees flow automatically into Google’s accounts. It doesn’t need roadshows, analyst pitches, or explanations to anyone about why it spends this money.

Google subsidizes AI tokens with ad profits, like someone with oil wells fighting a gas station’s prices—its oil is from its own land, while competitors’ oil is bought with bank loans.

OpenAI and Anthropic are those who buy oil on credit.

OpenAI has raised over $180 billion in total funding, with a valuation exceeding $850 billion. Anthropic has raised over $130 billion. These funds come from venture capital and strategic investors—they’re not charities; they expect these companies to go public and deliver hefty returns upon exit.

And once they go public, real trouble begins. Public companies must disclose financials to the world. Every quarter, Wall Street analysts scrutinize revenue, profit, customer acquisition costs, marginal costs. When they calculate that for every $1 of subscription revenue, the company actually loses $70—no matter how dazzling the growth story, the stock price can’t hold.

Bill Maris plainly states this logic on the podcast: “If I were Google, and decided to cut token prices by 80%, what would happen to OpenAI and Anthropic’s business models?”

The host asks how likely that is. Maris doesn’t hesitate: “100%. Capital as a weapon, tokens as a weapon.”

This isn’t just speculation. Bill Maris is the founder and CEO of Google Ventures, and a VP of special projects at Google, having incubated Waymo and Google X. Everyone present understands: this isn’t a hypothetical, he’s seen Google fight.

He describes a simple scenario: Google announces Gemini API prices cut by 80%. How would enterprise clients react? If the product quality is roughly the same—many benchmarks already show Gemini on par with Claude and GPT—would they keep paying more for the expensive one?

Maris’s own answer: “If you’re a company, and you can pay 80% less at Google or Gemini for essentially the same product, why wouldn’t you?”

And OpenAI and Anthropic have almost no countermeasures. They can’t follow suit with price cuts—no printing press, every dollar is from investors. They can’t maintain a premium through technical gaps—the difference between large models is shrinking rapidly; today you’re ahead by three months, tomorrow you’re caught up. It’s not like the iPhone vs. Nokia tech gap. The moat between AI models is more like a sandcastle—when the tide rises, it washes away.

Under Bill’s narrative, Google has a big advantage, but in the AI world, can Google truly monopolize? Meta can open-source a free model at any time, China has DeepSeek and ByteDance, Amazon is pushing its own models. When token prices drop to bargain levels, competitors don’t disappear—they also lower prices.

In the AI war, there may be no winners.

Token’s “infinite game”?

Even those unfamiliar with history can somewhat judge the current AI war’s outcome:

One is the “Internet Service” script—Didi’s story, Amazon’s story: subsidize first, monopolize, then raise prices to harvest. In this script, today’s price war is just the prelude; ultimately, one or two winners will dominate most of the market and gain pricing power. If so, the current huge losses are a worthwhile investment—just like Amazon lost money for twenty years before becoming a dual leader in e-commerce and cloud.

The second is the “utilities” script. Tokens become a standardized infrastructure resource—like electricity, bandwidth, cloud storage. No one can maintain long-term pricing power because product differences are tiny, switching costs are low. Competition drives prices down to cost, profit margins approach zero. Eventually, governments may step in to regulate—like they did a century ago with electricity and telecom.

The dividing line between these two scripts hinges on one word:

Lock-in.

Didi can raise prices because passengers are locked into the driver network, drivers are locked into order flow. Amazon can raise prices because merchants are locked into its logistics and traffic ecosystem.

Lock-in effect is the “burn first, profit later” foundation.

But AI tokens—already discussed repeatedly—almost lack lock-in. API standardization, switching costs near zero. The core condition for the first script to succeed doesn’t exist for tokens.

If the second script, the “utilities infrastructure” endgame, is more realistic, what we’re witnessing isn’t a war with a clear winner, but an endless consumption race.

Meituan founder Wang Xing once described this kind of competition. His insight: some competitions don’t have a “win” concept. Participants’ goal isn’t to beat opponents but to ensure they stay at the table. Because as long as they’re at the table, they can keep raising funds, hiring, iterating. Leaving the table is the only way to lose.

Re-examining today’s AI landscape through this lens, many seemingly contradictory phenomena suddenly become clear.

OpenAI’s latest valuation exceeds $800 billion, not because training models costs that much. It costs that much to keep fighting the price war. Fundraising isn’t about winning; it’s about “being qualified to keep fighting.”

Google plans to cut token prices by 80%, not to eliminate OpenAI and Anthropic. It’s to ensure it remains a core player in AI—just like it used free Android to stay at the top in the mobile era.

And Anthropic’s recent price hike for its flagship Fable 5 API—$10 per million input tokens, $50 per million output tokens—seems like a “price increase,” but is actually a strategic move to filter out only high-end enterprise clients willing to pay for advanced capabilities, because it knows: consumer subsidies can’t beat Google.

Every price war expands AI’s usage scale. Larger scale means more data, more scenarios, more developers entering the ecosystem. This, in turn, makes all models stronger. Participants attract resources through war—this isn’t a zero-sum game of life and death, but a process where everyone gets stronger through competition, yet none can earn huge profits.

Does this sound like the eventual shape of the electricity industry?

140 years ago, Edison and Westinghouse both thought they were fighting for a winner-takes-all market. They bet everything on “who defines the standard of electricity, owns the electricity.” But the fate of electricity teaches us a simple truth:

When a technology becomes sufficiently important, universal, and standardized, it no longer belongs to any single company. It becomes infrastructure.

The AI race, on the surface, is Google versus OpenAI versus Anthropic—model capabilities, funding scales. But in the bigger picture, this competition is accelerating the push of AI toward a foundational infrastructure level that no company can monopolize.

When Bill Maris says “it will 100% happen,” he might not just be predicting Google’s price cuts. He could be unconsciously foretelling a larger trend—within AI, tokens will ultimately belong to no one. Just like no one “owns” electricity today.

For OpenAI and Anthropic, this means an unsettling reality: even with technological leadership and massive funding, the future of “making big money from AI” might not exist from the start. They face not just a temporary price war, but a structural destiny—what they’re building might essentially be the next-generation water, electricity, and roads.

And for users, in some sense, that might be good news. As long as the subsidy war continues, people can still enjoy “good deals” at a $20 cost for $400 worth of computing power.