AI Giants' API Pricing Shifts Collectively, Subsidy Era Ends

After several years of exchanging low prices for market share, the top three global artificial intelligence model providers—Anthropic, OpenAI, and Google—are following distinctly different paths to shift their API pricing strategies from growth-first to profit-first, signaling an end to the subsidy era of this industry.

Google's Most Aggressive Price Increase

Google's Gemini product series has seen the most drastic price increase. Data shows that the annual compound growth rate (CAGR) of the Gemini Flash series on the input side is as high as 314%, and on the output side, it reaches 402%, far exceeding the United States' approximate 3% annual inflation rate. The Gemini Pro series follows a similar trajectory, with the output price having increased from less than $2 per million tokens in the early days to the current $12.

Nevertheless, Google's current flagship pricing—Gemini 3.1 Pro input $2, output $12 per million tokens—remains at about 40% of the competitors, maintaining a significant price advantage among the three main suppliers. Analysts believe this pricing strategy serves a dual purpose: on one hand, to avoid low-price competition from Chinese open-source models, and on the other hand, to gradually move closer to a profitable range.

OpenAI's Complex Script of Decrease Then Increase

OpenAI's pricing trajectory exhibits a typical strategic shift arc. The flagship model, from the era of GPT-4 with an output price as high as $60 per million tokens, experienced substantial price reductions through several generations of products such as GPT-4 Turbo and GPT-5, reaching a bottom, and then rebounding with the release of GPT-5.4 and GPT-5.5, with the current output price rising back to $30.

At the same time, the price for the budget-friendly Mini series, which targets users with budget concerns, moves in the opposite direction—the output CAGR reaches 44%, reflecting that OpenAI is simultaneously charging higher premiums across multiple market segments.

Market observers attribute this pattern to OpenAI's increasingly tense financial situation. The company has seen a sharp expansion in capital expenditures in recent years, establishing a user base with low prices during periods of abundant financing, and gradually regaining pricing power as cash pressures rise, which is a textbook example of converting market share into profit margins.

Anthropic Maintains Premium Positioning

In contrast, Anthropic's strategy is the most stable. The Claude Sonnet series has seen almost no change in price since 2023, with both input and output CAGRs approaching 0%, consistently maintaining a range of $3/$15 per million tokens, reflecting its consistent strategy of targeting the high-end enterprise-level secure AI market.

The price of the top-tier flagship Opus series on the input end has been reduced, with a CAGR of -16%, which analysts interpret as a tactical compromise in the face of increasing competition, rather than a strategic shift. The current pricing for Claude Opus 4.7 is $5 for input and $25 for output per million tokens, which is still among the highest in price among the three suppliers.

The Common Background of Capital Expenditure Pressure

Behind the collective shift in pricing strategies of the three companies is an unprecedented wave of capital expenditure. It is estimated that the AI infrastructure capital expenditures of Google, Microsoft, Amazon, and Meta will total more than $700 billion in 2026, a significant leap of over 70% from the approximately $410 billion in