Reducing AI Processing Costs Will Transform the Market

• New startups are reducing spending on AI processing requests.
• Open AI models may prove to be more expensive through higher spending on tokens.
• Lower spending on AI may expand access to technology for small businesses.

A team of developers from various nationalities conducts data analysis regarding marketing in social media. In today's business, spending on artificial intelligence (AI) can be significant: average spending on AI systems can reach up to $300 per month through subscriptions to various services, such as ChatGPT and Claude. Many do not realize that this spending is not only related to software subscriptions but also to hardware expenditures, which are used for processing requests.

Currently, new startups are emerging, such as Positron AI, Groq, Cerebras Systems, and Sambanova Systems, which aim to reduce spending on AI processing requests. If they manage to reduce these costs, current AI spending could decrease from $300 to $30. These companies are developing alternatives to expensive graphics processors from Nvidia, which currently dominate the market.

Research conducted by Nous Research revealed that open AI models consume significantly more compute resources than closed models, which may diminish their economic advantages. Specifically, open models can consume from 1.5 to 4 times more tokens for performing identical tasks, which can lead to significant costs.

These revelations may influence strategies for implementing AI in enterprises, as companies often focus on the accuracy of models and spending on tokens, ignoring general compute requirements. Closed models, such as those proposed by OpenAI, often demonstrate better efficiency in tokens, which can compensate for their higher price.

If the trend of reduced spending on AI continues, it may simplify access to powerful AI tools for small enterprises, freelancers, and educators, lowering entry barriers to the market.

Tags: Technology/AI