Artificial intelligence is becoming increasingly expensive. As a result, companies are beginning to rethink their rapid adoption of the technology.
After ChatGPT’s launch, AI firms followed a familiar Silicon Valley strategy. They offered services at very low prices to attract users quickly.
However, analysts now say that phase is ending. Kevin Simback of Delphi Labs describes it as an era of “subsidised intelligence”. In this phase, investors covered much of the cost.
He warns that the situation is changing. AI companies now face pressure to generate real revenue. In addition, firms like OpenAI and Anthropic are preparing for potential public listings.
Prices Rise as AI Agents Drive Costs
AI pricing is increasing across the industry. One major reason is the rise of AI agents.
Unlike basic chatbots, agents perform complex tasks. For example, they can book appointments, write code, or manage files.
However, these systems are expensive to operate. A single task may trigger multiple agents at once. As a result, costs increase quickly.
AI companies measure usage in tokens. One agent task can consume far more tokens than a simple chat.
Infrastructure Struggles to Keep Up
At the same time, demand for computing power is rising sharply.
Data centres and advanced chips are struggling to keep pace. Consequently, shortages are emerging across the industry.
Mark Barton of Omniux says costs are rising quickly in developer use cases. He adds that AI expenses are now increasing at an exponential rate.
“Tokenmaxxing” Drives Unexpected Bills
Some companies are overusing AI systems in what analysts call “tokenmaxxing”.
In some cases, firms deploy AI without strict controls. As a result, usage costs rise rapidly.
Jack Gold of J.Gold Associates says costs can exceed employee salaries within months. This happens when systems are used excessively.
Companies Rethink AI Strategy
Even major tech firms are adjusting their approach.
Meta previously encouraged employees to use AI tools freely. However, its CTO Andrew Bosworth later warned against unnecessary usage.
He said employees should not use AI “just for the sake of it”.
Similarly, Uber’s chief operating officer said AI spending has not yet delivered clear productivity gains. This comment has raised concerns in the industry.
Firms Turn to Cheaper AI Models
To reduce costs, companies are exploring new strategies.
Some are switching to open-source AI models. These tools are cheaper but less powerful than premium systems.
In addition, businesses are adopting smaller, specialised models. These are designed for specific industries such as finance or real estate.
Others are breaking tasks into smaller steps. Then, they assign each step to the cheapest suitable model.
Cost Differences Remain Huge
The price gap between models is significant.
According to consultants, large AI models can cost around $15 per million tokens. However, smaller models can reduce that cost to just a few cents.
As a result, companies are under pressure to optimise usage more carefully.
AI Moves Toward a Commodity Market
Experts say AI is gradually becoming a commodity. In this market, price matters as much as performance.
However, high-end systems are still important for advanced users. Some companies continue to pay for the most powerful models.
John Belton of Gabelli Funds says top-tier users will always pay for premium AI.
For now, the industry faces a balancing act. Companies must manage rising costs while still chasing innovation.
