AI's new frontier: OpenAI’s o3 redefines intelligence and economics

A recent Economist article sheds light on OpenAI's announcement of its new generative AI model, o3, just before Christmas. This development stirred a mix of excitement and scepticism. Enthusiasts viewed it as a potential leap toward superhuman intelligence, possibly surpassing the transformative impact of ChatGPT's launch in 2022. Critics, however, questioned OpenAI's decision not to release the model publicly, speculating it might be an effort to bolster its reputation under the incoming Trump administration.

One consensus has since emerged: o3 and its predecessor, o1, exhibit enhanced performance with more extensive computational "thinking." This increased computing power comes at a higher cost per query, signalling a shift in the economics of AI. Unlike traditional digital services with negligible marginal costs, advanced AI models demand significant resources, moving the industry away from the era of cheap, scalable internet services.

Valued at $157 billion, OpenAI is seen by investors as a rising tech giant. However, the high operational costs of cutting-edge models and competition from rivals may prevent it from enjoying the monopoly-like dominance of previous tech titans. François Chollet, a prominent AI researcher, highlighted this shift, describing o3 as a paradigm-altering development in AI economics.

OpenAI's o3 achieved a landmark score of 91.5% on Chollet's Abstraction and Reasoning Corpus (ARC), a set of visual reasoning challenges designed to be simple for humans but tough for AI. Unlike earlier models, o3 utilises a "test-time compute" approach, meaning it dedicates more computational effort during inference. While OpenAI’s initial attempt to solve ARC cost $6,677 (or $17 per question), the superior score came at an exorbitant cost, potentially $3,000 per query—underscoring the resource-intensive nature of this breakthrough.

This costlier approach disrupts traditional AI economics. With limited free access to powerful models, OpenAI now charges steep subscription fees, such as $200 per month for the pro version of o1 and potentially $2,000 monthly for o3. Meanwhile, suppliers like Nvidia and cloud providers such as Amazon, Microsoft, and Google stand to benefit as demand for computing power rises.

Competition further complicates OpenAI's dominance. Rivals like Google (with Gemini 2.0 Flash) and open-source alternatives are gaining traction. Analysts predict AI model development will resemble an oligopoly with specialised models catering to specific industries, rather than general-purpose tools like ChatGPT.

While o3 demonstrates OpenAI's technical prowess, its long-term business success hinges on how effectively it creates value to justify its steep costs.

By Vugar Khalilov