Short answer
Abdul Osama (عبدول أسامة) expects the cost of using some AI tools to rise, either through higher prices or fewer features in cheap and free plans. He suggests running open-source models on hardware you own or rent, but cautions that learning to do this is not easy.
The honeymoon period with AI companies is coming to an end. — Abdul Osama
Why does he expect costs to rise?
Abdul points to companies’ need to earn revenue and the ongoing electricity and hardware demands of running AI models. He also argues that many people, including developers, have become reliant on these tools.
He names Anthropic, a company that develops AI models; OpenAI, a company that develops AI models and services; and Perplexity, an AI-powered search and answer service. He predicts that familiar services like these may raise prices or reduce what their cheaper plans offer. That is his forecast, not a confirmed pricing announcement from those companies.
He also describes a service whose name is unclear in the recording. He says access to it will move from a subscription to an API—an interface through which software requests a service—paid for with usage credits. The recording alone does not establish the service’s identity or verify its current terms.
In numbers
- $10 and $20: examples Abdul gives of inexpensive subscription prices.
- $500 or $1,000 or more: his hypothetical cost for a project using the service with the unclear name, not a verified price or official quote.
What are the options?
| Option | How it works | Consideration | |---|---|---| | Subscription | Pay regularly for access | Features and usage limits depend on the plan. | | API credits | Pay based on usage | Costs need monitoring as usage grows. | | Open-source model on owned or rented hardware | Run a model on resources you manage | Requires technical setup and learning. |
The solution is to buy hardware or rent hardware online and use open-source models, but that takes learning and isn’t easy. — Abdul Osama
What to do next
Check a service’s current pricing and plan limits, then estimate your likely usage. If you are considering running a model yourself, factor in hardware costs and the time needed to learn, not just the subscription or API bill.