Paying for AI? ChatGPT and Claude May Soon Cost Less
Something profound has changed in the artificial intelligence arms race. For three years, we have grown used to leading companies competing over abstract titles such as “the smartest model,” “the most creative,” or the one capable of solving the most complex physics equations. But as a report in The Wall Street Journal suggests, the real battleground is now shifting from the labs to the profit and loss statements of business customers, and from there to individual consumers as well. According to the report, OpenAI is considering an aggressive price cut aimed at slowing the meteoric rise of its bitter rival, Anthropic.
The backdrop to the move is also a growing sense of exhaustion in the business sector over the inflated bills coming from technology companies, especially AI firms. The professional term that has gained popularity in the industry in recent months is “Tokenmaxting,” meaning maximizing the price of tokens, the units used to measure information volume and billing in AI systems. The term describes a phenomenon in which organizations consume enormous amounts of computing resources and tokens without always seeing a clear return on investment. Around the world, open frustration with this trend has begun to surface, and major companies such as Uber have even acknowledged that their annual budgets for AI-based code development were completely exhausted as early as April. Even Sam Altman, OpenAI’s chief executive, has been forced to admit in closed events that computing costs have become a huge problem for users, and that the company must find ways to deliver more value for less money.
But the real pressure on the maker of ChatGPT does not stem only from customer complaints, it comes mainly from the existential threat posed by Anthropic. The young company, founded by OpenAI defectors, has seen phenomenal success over the past year and a half thanks to its dedicated code development tool, Claude Code, and more recently with the launch of its new model, Pable 5. These tools have gone viral among software engineers, pushing Anthropic’s annual revenue run rate to unprecedented levels that are rapidly narrowing the gap with OpenAI. In several private funding rounds, Anthropic’s estimated market value has even surpassed OpenAI’s for the first time, prompting the latter to shift its management focus to the urgent promotion of its competing programming tool, Codex.
The current move to lower token prices creates a huge financial paradox. Both companies are now on a direct path toward massive initial public offerings on Wall Street, and a decision to deliberately squeeze profit margins just before going public is considered a bold, if not risky, strategy. OpenAI is burning through cash at a rate 14 times faster than Anthropic, and its current forecasts do not project a move to profitability or positive cash flow before 2030. Launching a price war at this stage could deepen the losses of both companies, which together are already spending tens of billions of dollars a year just on maintaining server farms and purchasing advanced chips.
Historically and technologically, this price war marks the stage at which a coveted premium product turns into a basic commodity measured by price rather than uniqueness. A similar process took place in cloud computing, when Amazon, Microsoft, and Google waged brutal wars of attrition over storage and processing prices until the market stabilized. As the quality gaps between different models narrow, and the ability of customers to switch from one system to another becomes simple and nearly effortless, companies are realizing that control of the market will no longer be determined by the machine’s level of intelligence, but by the ability to offer it at the lowest price. The big winner, at least in the short term, will be the industry as a whole, which will benefit from far cheaper artificial intelligence, but for investors who must navigate the upcoming offerings, it is a nerve-racking bet.
