At a time when AI companies are collectively losing tens of billions of dollars due to enormous computing infrastructure costs, substantial price cuts would further erode profit margins. Why would OpenAI choose to compress its revenue stream at this particular moment? Simply put, there are three key reasons.

The Rise of Claude Code and Growing Market Share Pressure

Anthropic is rapidly closing the gap from its position as the number-two player.

Over the past year, Claude has achieved significant growth in both the developer and enterprise markets. According to Similarweb data, Claude’s share of global generative AI website traffic increased from 6.0% to 8.9%. While the absolute figure remains relatively modest, its growth rate is the fastest among major AI platforms.

More importantly, Anthropic’s enterprise revenue is projected to reach US$10.9 billion in the second quarter, representing growth of more than 130%, while generating operating profit of US$559 million.

Meanwhile, OpenAI continues to operate at a loss, reportedly losing approximately US$1.22 for every US$1 of revenue generated.

Claude Code has become the key variable behind Anthropic’s rise.

The product has gained tremendous popularity among software engineers, prompting many companies to redirect portions of their AI budgets toward Anthropic. Some firms have reportedly exhausted their entire 2026 AI budgets within just four months.

According to media reports, one of the primary reasons behind Anthropic’s revenue surge is Claude Code itself. Programmers are precisely the type of users most willing to pay. A US$100 subscription fee can potentially save companies thousands of dollars in engineering costs.

OpenAI’s valuation currently stands at approximately US$852 billion, having been surpassed by Anthropic’s US$965 billion valuation.

This helps explain the direct motivation behind a potential pricing war.

OpenAI needs to re-establish its position in the highly valuable enterprise coding-assistant market, and lowering Token prices represents the most direct customer-acquisition signal available.

The Subscription Model Is Becoming Unsustainable

On June 1, 2026, Microsoft announced that all GitHub Copilot plans would officially transition to a Token-based pricing model, with the standard US$19 monthly subscription converted into an equivalent Token allowance.

This shift exposed the true costs that subscription pricing had long concealed.

According to user calculations, a single AI-assisted coding session can consume between US$30 and US$40 worth of Tokens, meaning an entire monthly subscription allowance may be exhausted in just one use case.

SemiAnalysis provided even more striking figures through a stress test of subscription economics.

For ChatGPT Pro, which costs US$200 per month, gross margins reportedly turn negative once average utilization exceeds 5.7%.

At 100% utilization, gross margins could theoretically fall to -1650%.

Heavy users running coding tasks and long-context workflows can rapidly consume their subscriptions, causing profitability to deteriorate sharply.

Enterprise Token bills are already exploding.

Uber reportedly exhausted its entire annual AI budget during the first four months of 2026.

Company executives publicly stated that a clear relationship between rising Token consumption and meaningful productivity improvements “has not yet been established.”

They even coined a new term:

“Tokenmaxxing”

The term describes employees generating Token usage simply to increase activity, even when the underlying tasks provide little practical value.

Data from Vercel’s AI Gateway shows that DeepSeek’s share of Token traffic surged from less than 1% to 17% within a single month, surpassing OpenAI’s 13%.

Yet because DeepSeek’s pricing remains extremely low, total customer spending on DeepSeek accounts for only around 1% of the gateway’s overall expenditure.

This suggests that enterprise customers are already actively searching for lower-cost alternatives, and DeepSeek has become one of the primary beneficiaries of this migration.

OpenAI CEO Sam Altman has publicly acknowledged the issue.

At a recent event, Altman admitted that AI usage costs have become “a huge problem” and stated that the company intends to help customers “get more value for less money.”

Pricing Strategy Ahead of IPOs

OpenAI confidentially filed for an IPO on June 8.

Anthropic reportedly submitted its own filing on June 1.

Both companies are now approaching highly anticipated public listings, and any pricing war will effectively become an early stress test of their business models.

One of the key risks identified by investors is the relatively high substitutability between the two platforms. Customers face relatively low switching costs when moving from one provider to another.

Both companies face the same fundamental dilemma:

Users want lower costs, while shareholders want profitability.

OpenAI has reportedly committed more than US$1.4 trillion to future computing and infrastructure investments.

Inference costs alone are expected to reach approximately US$14.1 billion in 2026, consuming nearly all of the company’s revenue.

Operating losses for 2026 are projected to reach US$14 billion, while cash burn is estimated at roughly US$1.7 billion per month.

Under such a financial structure, lowering prices would undoubtedly place even greater pressure on profitability.

However, failing to offer more competitive pricing creates its own risks.

In IPO documents submitted earlier this week, OpenAI acknowledged that: “Some things may be more convenient to do as a private company.”

The statement appears to reference the fact that certain competitive strategies and business adjustments become more difficult under the disclosure requirements imposed by public markets.

Whether a lower-pricing strategy can successfully increase market share without materially expanding losses will become one of the most important questions investors evaluate when assessing OpenAI’s future valuation as a public company.