AI Inference Costs - What Happens When Subsidies End
Basically, AI might get more expensive for advanced tasks while remaining cheap for everyday use.
AI inference costs are on the rise as subsidies fade. Major labs like OpenAI face financial challenges, leading to a split in AI pricing. While advanced models may become costly, everyday tasks will likely remain affordable.
What Happened
The era of subsidized AI inference is coming to an end. Major labs like OpenAI and Anthropic are currently losing money on inference, and this financial strain is expected to lead to significant changes in the AI landscape. As these companies can no longer afford to offer services at a loss, the costs associated with high-end AI models will likely rise. This shift will create a divide between expensive frontier models and more affordable options for everyday users.
Who's Affected
The changes in AI inference costs will impact both consumers and businesses. For consumers, the most common AI tasks—like writing summaries or answering questions—are expected to remain affordable. However, businesses relying on cutting-edge AI for complex tasks may face increased expenses. The divide means that while the majority of users may not notice a difference, those requiring advanced capabilities will need to budget for higher costs.
What Data Was Exposed
Recent reports indicate that OpenAI's revenue could reach $13.1 billion in 2025, yet they project a staggering cumulative cash burn of $115 billion by 2029. This financial outlook underscores the unsustainable nature of current subsidy practices. Additionally, the decline in inference costs has been dramatic, with a reported 280-fold decrease in GPT-3.5-level inference costs documented by Stanford HAI. As subsidy support wanes, the market will have to adapt to these new financial realities.
What You Should Do
For businesses and individuals, it’s essential to prepare for the upcoming changes in AI costs. Here are some steps to consider:
- Evaluate your AI needs: Determine whether your tasks require high-end models or if more affordable options will suffice.
- Stay informed: Keep an eye on developments in open-source AI models, which may provide cost-effective alternatives.
- Budget accordingly: As costs rise for advanced AI services, ensure your financial plans reflect these changes.
By understanding the implications of ending subsidies, users can make informed decisions about their AI usage moving forward.