TLDR: A recent article from The Conversation Canada explores the controversial idea that generative AI (genAI) might ultimately prove to be economically worthless, despite the current hype. Author Fenwick McKelvey suggests that this outcome, driven by high operational costs, minimal productivity gains, and significant legal liabilities like copyright infringement, could actually have positive implications. If genAI fails to generate sustainable profits, it could lead to a stall in its development, leaving consumers with free, ‘good enough’ tools and reducing the influence of large tech firms. The article highlights that the true value of knowledge, on which genAI models are trained, may be immeasurable, making it difficult to truly ‘own’ or commercialize these systems.
In the current rush to capitalize on the generative artificial intelligence (genAI) gold rush, a less-discussed possibility is emerging: what if the technology never performs well enough to replace human workers, companies struggle to implement AI effectively, or most AI startups simply fail? This scenario, explored by Fenwick McKelvey of Concordia University in an article for The Conversation Canada, suggests that such an outcome might not be a negative development.
Current estimates indicate that major AI firms are facing a substantial US$800 billion revenue shortfall. The productivity gains from genAI have so far been minimal, primarily benefiting programmers and copywriters. While genAI offers some useful functionalities, it has not yet become the driving force of a new economy.
This alternative future, though different from the one dominating headlines, could be perfectly acceptable. It challenges the narrative promoted by AI firms, which often relies on hype to attract investment and promise massive future profits. McKelvey posits that if genAI turns out to be economically worthless, it might actually be a positive development.
Indispensable or Indefensible? The Economics of GenAI
Free and low-cost subscription genAI services, such as ChatGPT and Gemini, incur significant operational expenses. Questions are increasingly being raised about how AI firms intend to generate revenue. OpenAI CEO Sam Altman has openly acknowledged the high costs, once humorously noting that every ‘please’ or ‘thank you’ from ChatGPT costs millions. He has also indicated that even paid ‘pro’ accounts are unprofitable due to the extensive computing power required for each query.
GenAI firms have largely followed the traditional startup model of burning through capital to attract and retain users with a compelling product. However, most successful tech giants have thrived by offering low-cost, indispensable products, often supported by advertising. When companies attempt to extract new value, it can lead to what journalist Cory Doctorow terms ‘enshittification’ – a gradual decline in platform quality, often marked by an increase in advertisements to offset losses from free services. OpenAI is reportedly considering introducing ads to ChatGPT, though with a cautious approach.
It remains uncertain whether this strategy will prove successful for genAI. There is a possibility that advertising revenue may not be sufficient to justify the immense operational costs, especially as genAI increasingly presents itself as a liability.
The Hidden Costs and Legal Liabilities of AI Models
Another significant challenge for genAI is copyright. Many AI firms are currently facing lawsuits for using copyrighted content without permission or are entering into expensive licensing agreements. GenAI models have been trained on vast amounts of data, including copyrighted books and scraped online content. One model, for instance, can reportedly recall 42% of the first Harry Potter novel from memory.
Firms are grappling with the substantial financial burden of lobbying for copyright exemptions and negotiating payments with publishers and creators to safeguard their models. These models, regardless of legal outcomes, could ultimately become significant liabilities. For example, American AI startup Anthropic’s proposed US$1.5 billion settlement to pay authors approximately US$3,000 per book for training its models was rejected by courts for being overly simplistic. Anthropic’s current valuation of US$183 billion could rapidly diminish under the weight of such lawsuits.
The cumulative effect of these factors suggests that AI may simply be too expensive to be ‘owned,’ transforming it into a ‘toxic asset’ – something useful in principle but lacking inherent financial value.
The Rise of Cheap or Free GenAI
Meta has adopted a strategic approach by releasing its genAI model, Llama, as open source. This move, whether intended to disrupt competitors or signal a different ethical stance, allows anyone with a capable computer to run a local version of Llama for free.
Open AI models represent another corporate strategy aimed at securing market share, with intriguing side effects. While not as advanced as commercial models like Gemini or ChatGPT, they are ‘good enough’ and are either free or significantly cheaper. These open models challenge the high valuations placed on commercial AI firms. The release of an open model by Chinese firm DeepSeek, which matched the performance of commercial models, temporarily caused a dip in AI stocks and fueled doubts about genAI’s perceived value.
Open models, emerging as a consequence of industrial competition, are becoming widespread and more accessible. If they continue to gain traction, commercial AI firms may struggle to sell their services against free alternatives. This could also lead to increased investor skepticism, potentially drying up funding for commercial AI ventures. Even if open-access models face legal challenges and are eventually removed from the internet, their widespread distribution would make complete eradication difficult.
Can Knowledge Truly Be Owned?
The notion of genAI becoming worthless may underscore the intangible value of knowledge itself. The most sophisticated genAI models are trained on the entirety of the world’s knowledge – an immense volume of information whose true cost is arguably incalculable. Ironically, the very efforts by AI firms to capture and commercialize this collective knowledge might be the undoing of their products. A resource so profoundly valuable that a price cannot be attached to it suggests that the outputs of these systems, being so deeply indebted to collective intellectual labor, may not be genuinely ‘ownable.’
Also Read:
- Massive Investments Fuel Generative AI Sector, Raising Bubble Concerns
- Generative AI: Reshaping Sales, Marketing, and Business Automation in 2025 and Beyond
If genAI proves incapable of generating sustainable profits, the consequences will likely be varied. Creators seeking deals with AI firms may find themselves out of luck, as major players like OpenAI, Anthropic, or Google would be unwilling to issue large payments if their models are financial liabilities. The progress of genAI could also stagnate, leaving consumers with functional, ‘good enough’ tools that are free to use. In this scenario, AI firms might become less influential, and the technology itself less dominant – an outcome that could be perfectly acceptable. Users would still benefit from accessible tools, while being spared from another cycle of overhyped promises destined to fail. The potential for AI to be less valuable than anticipated could serve as a crucial safeguard against the escalating power of today’s big tech companies. If the business case for generative AI proves unsustainable, its financial downfall could be the most effective check on its burgeoning influence.
