Two new reports set out a complicated future for artificial intelligence (AI) in the immediate years ahead.
A new report by non-profit research institute Epoch AI casts doubt on the notion that AI companies, including OpenAI, will soon turn consistent profits.
Separately, business and technology insights company Gartner predicts that by 2027, 35% of countries will be locked into region-specific AI platforms using proprietary contextual data. Platform lock-in looks set to rise from 5% to 35% by 2027.
On profitability, Epoch AI findings suggest that while running large language models (LLMs) generates significant revenue, the cost of developing the next generation outweighs any profit. This research by Jaime Sevilla, Hannah Petrovic, and Anson Ho analyzed OpenAI’s “GPT-5 bundle” which, over its lifetime, generated revenue estimated at $6.1 billion. However, operational costs—including $3.2 billion (€2.7 billion) for inference computing, $1.2 billion (€1.0 billion) in staff compensation, $2.2 billion (€1.8 billion) in marketing, and $0.2 billion (€0.17 billion) in legal and administrative expenses—left slim margins. While gross profits of $2.9 billion (€2.4 billion) suggest AI models are profitable to run, the short lifecycle of each model makes recouping the cost of developing the next big model difficult.
Despite lifecycle losses on individual models, AI products do not need to be profitable immediately, as long as companies can generate revenue over time. Falling computing costs, stickier enterprise contracts, and longer model relevance can improve financial performance.
Slower model releases, diversified offerings, and integrating AI into other software are approaches being used to manage costs and revenue streams.
Meanwhile the rise of a more nationally focused form of AI delivery is influencing a segmented world market. According to Gaurav Gupta, VP Analyst at Gartner
Countries with digital sovereignty goals are increasing investment in domestic AI stacks as they look for alternatives to the closed U.S. model, including computing power, data centers, infrastructure and models aligned with local laws, culture and region.
The report predicts that “nations establishing a sovereign AI stack will need to spend at least 1% of their GDP on AI infrastructure by 2029.” Such developments echo the thoughts of europeanconservative.com’s Carlos Perona Calvete, who called for
An optimistic model for integrating AI in ways that do not subordinate local needs and cultures, but instead empowers them, could constitute a major plank of future political platforms.
