For several years, AI pricing has largely followed a predictable path, with new models arriving faster, smarter, and often cheaper or at most slightly more expensive than their predecessors.
Budget-tier models, designed for “good-enough” intelligence at minimal cost, were always assumed to maintain stable pricing or even become more affordable over time. Google’s Flash, OpenAI’s mini and nano, and Anthropic’s Haiku exemplified this approach.
That assumption was upended with the release of Google’s Gemini 3.5 Flash on May 19. The new model is priced at $1.50 per million input tokens and $9 per million output tokens—triple the cost of the Gemini 3 Flash, which charged $0.50 and $3 respectively. This sharp increase in what was traditionally the low-cost tier signals the end of an era of subsidized AI access.
Experts caution against misreading the change as a rise in the intrinsic cost of AI. The technological efficiency and cost of producing a given level of intelligence continue to decline rapidly. What has shifted is the approach of major AI labs: rather than passing on these savings to consumers, they are now quietly exploring how much customers are willing to pay. In contrast, DeepSeek has deliberately kept its pricing low, attracting users and demonstrating that affordability remains a competitive advantage.
The development marks a broader turning point in the AI industry. As the sector matures, the economics of AI are no longer solely about technological capability but increasingly about market strategy and consumer willingness to absorb higher costs. Analysts predict that this new pricing dynamic will shape AI adoption, accessibility, and the competitive landscape for years to come.