The commercial narrative for generative AI is undergoing its most profound self-examination in three years. The industry has rapidly progressed through three distinct commercial phases—from user acquisition via subsidies, to monthly subscriptions that hid costs, to per-Token billing that has triggered a corporate invoice crisis. A potential price war could now reset this entire monetization logic.
According to a report, OpenAI is considering significantly reducing the fees it charges users per Token, aiming to win enterprise clients from competitor Anthropic. Sources indicate this move is partly preemptive, as OpenAI anticipates Anthropic will take similar price-cutting actions. OpenAI CEO Sam Altman recently acknowledged that AI usage costs have become "a huge problem" and stated the company would "help people get more value for less money."
The timing of this news is particularly sensitive. OpenAI has confidentially filed for an IPO this week, and Anthropic is also in the final countdown to its own public listing. Concurrently, a key industry index tracking LLM Token expenditure has fallen for seven consecutive trading sessions, its longest losing streak since January, reflecting deep market anxiety over the sustainability of AI bills. The report states that a price war would directly erode the profit margins of both companies, which are already losing billions annually due to the massive compute power required for their AI systems.
The core of this discussion is no longer just a single pricing decision, but a more fundamental question: as the narrative of "more Token consumption is better" reaches its end, who will tell the AI industry's next commercial story, and how will it be told?
The Three-Phase Evolution: From Subsidies to Token Bills
Generative AI's commercialization has undergone three clear stages in just three years.
The first stage established the industry tone with monthly and annual subscriptions. In February 2023, OpenAI launched ChatGPT Plus at $19.99 per month, pioneering consumer-paid access to large language models. Other tech giants followed, making fixed-fee subscriptions the standard for initial business models.
The second stage saw the full eruption of a subsidy war. To boost Annual Recurring Revenue (ARR), a key metric for funding valuations, vendors turned to large-scale subsidies. Alphabet offered students free access to Gemini Advanced for 15 months, OpenAI introduced a Team membership at $1 for the first month, and other companies entered with deeply discounted pricing. The essence of subsidies was trading losses for growth, with reports indicating significant per-user losses on some subscription services.
The third stage was the forced switch to usage-based billing. The announcement that all plans for a major coding tool would formally shift to Token-based consumption on June 1, 2026, brought the true costs long-hidden by subscriptions into full view. User calculations revealed that a single intensive session could consume a month's worth of Token credits.
Bill Shock: When Tokens Cost More Than People
The implementation of per-Token billing has laid bare the true scale of corporate AI expenditure.
Enterprise bill figures are staggering. A top executive at a major ride-sharing firm publicly stated in May 2026 that the line between increased Token consumption and substantive product improvement "doesn't exist yet," coining the term "tokenmaxxing" to describe employees performing valueless tasks to boost usage metrics.
More direct data shows one company exhausted its annual Token budget within the first four months of 2026, while another forecasts its annual payments to Anthropic will reach approximately $3 billion.
Anthropic's own developer documentation suggests the average daily cost for a user of its coding tool is around $13, with 90% of users spending under $30 per day. This translates to potential annual Token costs exceeding $75,000 for a small development team.
The return on investment is equally alarming. An analysis of data from 2,444 companies found that for every dollar spent on AI Tokens, only 18 cents generated actual user-facing value. The rest was consumed by fixing AI-introduced bugs, rework, and review friction.
Faced with runaway bills, enterprises are beginning to actively control usage. One tech giant halted internal AI usage leaderboards, instructing employees not to use AI for its own sake. Another plans to gradually phase out subscriptions to a coding tool for some employees. Analysts note that AI Token spending for some firms already accounts for 10% of total employee labor costs, a figure that may rise further.
The Fourth Act: Price War Ignites as OpenAI Considers Deep Cuts
It is against this backdrop that the fuse for a price war has been lit.
Reports indicate Altman's consideration of price cuts is directly triggered by the pressure of catching up to Anthropic. Anthropic's revenue has surged recently, driven by the popularity of its Claude Code tool among software engineers, and the five-year-old startup's valuation has, for the first time, surpassed that of OpenAI.
However, the cost of this price war will be exceptionally heavy. Significant price reductions would further compress the already negative profit margins of both companies within a fiercely competitive landscape.
A fundamental risk long identified by investors is the high substitutability between OpenAI's and Anthropic's products, allowing clients to switch easily. This means price cuts may temporarily retain customers but fail to build a true moat, merely delaying market share erosion.
This dilemma is also transmitted outward through the financial loop between cloud computing giants and AI labs. OpenAI and Anthropic together account for over half of the future cloud service commitments from major providers. If price cuts lead to downward revenue revisions, this entire chain could face pressure.
An expert in the field commented that this further exposes OpenAI's fragility and the severity of its predicament, with potential ripple effects across the tech supply chain.
A divergence of views is playing out on Wall Street. One analyst from a major bank views current bill anxiety as merely a "minimum speed bump on the road to higher spending," arguing that lower per-Token prices coupled with rising adoption and the higher consumption of agentic AI will lead to significantly higher total long-term expenditure.
A more pessimistic view from another analyst suggests the current industry boom has funneled nearly all value to semiconductor companies, a phenomenon deemed "historically unprecedented and unsustainable." Once enterprises face the true price of usage-based billing, the capital flows supporting GPU purchases and model training could reverse.
The Fifth Act: What's the Next Chapter for Token Economics?
The next chapter of AI industry commercialization post-price war remains unwritten, but its outline is emerging.
One framework suggests a move towards tiered pricing based on scarcity. The core logic is that compute-intensive frontier AI will not disappear but will increasingly concentrate in the hands of a few large enterprises that can bear the cost. For the broader market, simpler models may be a more productive path until physical constraints ease, leading to a stratified AI usage landscape.
The more optimistic bank judgment holds that even if per-Token prices fall, the proliferation of agentic AI could multiply Token consumption per task by several times, potentially still expanding overall spending. The current bill anxiety might just be a temporary slowdown.
The concept of "valuemaxxing" has been proposed, advocating for the industry to shift from maximizing Token consumption to maximizing the value generated per Token. While this direction is gaining consensus, true commercial implementation requires AI labs to find a pricing system that reflects real costs while remaining acceptable to enterprise clients—the core unresolved issue in the current debate.
However, the most overlooked variable in this price war may be Chinese AI models. Recent data from a U.S. corporate spend management platform shows a Chinese model topping the growth chart for U.S. enterprise software subscriptions. The platform's chief economist emphasized this represents real, paid API usage, not local deployment of open-source models, admitting surprise at U.S. companies adopting it. Third-party estimates suggest the API pricing for this leading Chinese model is roughly one-tenth that of a major U.S. model.
As OpenAI and Anthropic battle, the ultimate beneficiary might be a player that has embedded "affordable pricing" into its DNA and isn't beholden to IPO investors for profitability. This may not be the most desired outcome of the price war, but it is becoming an increasingly unavoidable reality.
Comments