Will AI Costs Soon Be Borne by Users?
AI is at the center of attention: integrated into an exponentially growing number of services, this technology has more and more users. But its extremely high costs compromise its profitability. Will the decision to make consumers pay (in)directly be part of Tech giants' strategies in the future?
March 7, 2025
Training and making AI systems available to administrations, businesses, and citizens: the operation has proven to be both extremely costly and unprofitable for Tech giants so far.
$5 billion
that's the amount of OpenAI's losses in 2024. Its revenue amounted to $3.7 billion.
Source: TechCrunch
In late February 2025, Microsoft (OpenAI's main investor) restricted its data center-related expenses by canceling several of its lease contracts. Yet these infrastructures are at the heart of how artificial intelligences operate, both during training and use phases.
But their cost proves to be extremely high: the required electronic chips are particularly expensive, while data centers continuously consume water for machine cooling as well as electricity for their operation.
For now, the revenue generated by AI tools is not enough to make these technologies profitable services. Sam Altman, OpenAI's CEO, even admitted that the company is losing money with its paid ChatGPT Pro offering, sold at $200 per month, because subscribers are using it more than expected. Yet a query launched on one of the company's most advanced models can now cost it $1,000.
Will consumers be progressively forced to bear AI operating costs?
Microsoft's current problem is having massively equipped its services with its Copilot AI. Having already invested billions in it, the company is now seeking to reduce its costs. The strategic decision not to operate additional data centers goes in this direction, as does increasing the price of its licenses.
30 to 45%
increase on Microsoft 365 subscriptions for individuals has been applied in the United States since January 2025.
Source: CNBC
Another avenue Microsoft could explore, in addition to adding advertisements to its AI-equipped services, would be to have them run directly from users' devices. A way to make citizens bear the associated operating costs (particularly in terms of electricity consumption).
A shift that would then require them to invest in increasingly powerful electronic equipment to keep up with the evolution of AI programs developed by Tech giants. This would accelerate the consumption of polluting devices at all stages of their lifecycle, and further deplete the natural resources exploited for their manufacture.
Beyond these environmental issues, such a strategy could also shape a multi-speed society. On one side, users who can afford devices capable of running the most powerful AI models... on the other, people forced to use less powerful tools for lack of being able to equip themselves with a state-of-the-art machine. A societal model evoking a true dystopia.
References:
- TechCrunch - OpenAI is losing money on its pricey ChatGPT Pro plan, CEO Sam Altman says
- The Conversation - Microsoft cuts data centre plans and hikes prices in push to make users carry AI costs
- CNBC - Microsoft raises price of consumer version of Microsoft 365 for first time in 12 years
[Cover photo: Esra Afşar]