 |
. | The cost of AI is significantly decreasing, with expenses for running AI models dropping dramatically, making it more accessible for startups and smaller companies. This shift is leading to a new landscape where the question is no longer if companies can afford to use AI, but rather if they can afford not to use it. |
The AI cost collapse is changing what’s possible for startups
The hidden costs of AI: How generative models are reshaping corporate budgets
Current Trends in AI Costs
Significant Cost Reductions
-
The cost of using AI has dramatically decreased over the past year. For instance, the cost of running 500 GPT-4 queries per user session has dropped from $5 to less than 50 cents.
-
OpenAI's token costs have fallen by 90%, making AI more accessible to smaller companies and startups.
Factors Contributing to Cost Changes
-
Model Efficiency: Smaller AI models can now perform nearly as well as larger ones when trained effectively. This has led to reduced costs for companies developing AI solutions.
-
Self-Generating Training Data: Companies are now using existing models to create their own training data, cutting data costs by up to 90%.
-
Improved Inference: AI models have become more efficient, allowing them to run on less powerful hardware, such as laptops, rather than requiring high-end GPUs.
Future Cost Projections
-
Despite current reductions, the average cost of computing is expected to rise by 89% between 2023 and 2025, primarily driven by the demand for generative AI.
-
Companies are exploring hybrid cloud architectures to manage rising costs effectively, ensuring they can scale AI without prohibitive expenses.
Energy Consumption Concerns
- The increasing use of AI is leading to higher electricity consumption. AI data centers are projected to consume 6% of the U.S. electricity by 2026, raising concerns about the sustainability of this growth.
These trends indicate a complex landscape where AI costs are decreasing in some areas while rising in others, particularly regarding infrastructure and energy needs.
No comments:
Post a Comment