In the rush to capitalize on generative artificial intelligence (GenAI), one possible outcome that is seldom discussed is the technology’s potential failure to effectively replace workers, leading to company mismanagement of AI or the collapse of many startups. Current estimates indicate that major AI companies may face an $800 billion shortfall in revenue. To date, the productivity gains from GenAI have been limited, primarily benefiting programmers and copywriters. While GenAI offers some impressive features, it has yet to become the driving force of a new economy. This outlook diverges from the narrative that AI companies prefer to promote. The hype surrounding GenAI continues to attract investments, promising significant future profits.
However, it is possible that GenAI may ultimately prove to be worthless, which may not be a negative outcome. The question of whether GenAI services are indispensable or indefensible arises, particularly as free and low-cost subscription services like ChatGPT and Gemini incur substantial operational costs. Currently, there are increasing concerns regarding how AI companies can achieve profitability. OpenAI’s CEO Sam Altman has openly discussed the substantial expenses incurred by his company, humorously noting that every polite response from ChatGPT can cost millions. The exact losses OpenAI faces per interaction remain uncertain, but even premium subscriptions reportedly operate at a loss due to high computational expenses.
Like many startups, GenAI firms have adhered to a traditional strategy: spending heavily to attract users with an irresistible product. However, successful tech companies have typically achieved success by offering low-cost products that users find indispensable, often funded through advertising. As companies seek new value, they may experience what journalist Cory Doctorow describes as ‘enshittification,’ characterized by a gradual decline in service quality. In this scenario, the increasing number of advertisements compensates for the loss of free offerings. OpenAI is contemplating introducing ads to ChatGPT, albeit with a commitment to being ‘thoughtful and tasteful’ in the approach. It remains uncertain whether this strategy will succeed for GenAI.
There is a risk that advertising may not generate sufficient revenue to cover the substantial expenses involved in its operation, as GenAI is evolving into a potential liability. Another significant issue for GenAI is copyright. Many AI companies are currently facing lawsuits for unauthorized use of content or are negotiating costly licensing agreements. GenAI has learned from numerous questionable sources, including copyrighted materials and online content scraping. For instance, one model can recall 42 percent of the text from the first Harry Potter novel. Companies are faced with the daunting financial burden of lobbying for copyright exemptions and compensating publishers and creators to safeguard their models, which could eventually prove to be a liability.
American AI startup Anthropic attempted to compensate authors approximately $3,000 per book to train its models, resulting in a proposed settlement of $1.5 billion, which was promptly rejected by the courts for its simplicity. Anthropic’s current valuation of $183 billion could quickly diminish due to ongoing litigation. Consequently, AI is becoming prohibitively expensive to manage, resembling a toxic asset—something that is useful but lacks intrinsic value. Meta has strategically released its GenAI model, Llama, as open-source, allowing anyone with a decent computer to run their own local version for free. This move may have been intended to disrupt competitors or to signal a different ethical stance, as it undermines the high valuations of AI firms.
Although OpenAI’s models are also aimed at securing market share, they are not as advanced as Gemini or ChatGPT, yet they are sufficiently functional and more affordable. The introduction of open models disrupts the high valuations attributed to AI companies. Chinese firm DeepSeek caused a temporary decline in AI stocks when it launched an open model that matched the performance of commercial offerings. While DeepSeek’s motives remain unclear, its success raises growing doubts about the perceived value of GenAI. The accessibility of open models is increasing, and as they gain traction, commercial AI firms may struggle to compete with free alternatives. This scenario could lead to investor skepticism regarding commercial AI, potentially resulting in a reduction in funding.
Even if open access models face litigation, removing them from the internet will be significantly more challenging. The notion of GenAI being worthless may highlight the idea that knowledge is inherently valuable. The best GenAI models are built upon vast amounts of global knowledge, making it difficult to determine a true price. Ironically, AI companies’ attempts to monetize this collective knowledge could ultimately hinder their products, as the resource is so valuable that assigning a price may be unfeasible. If GenAI fails to produce sustainable profits, the outcomes could be mixed. Creators seeking partnerships with AI firms could be disappointed, as there may be no substantial payments from OpenAI, Anthropic, or Google if their models prove to be liabilities.
Progress in GenAI could stagnate, leaving users with merely ‘good enough’ tools that are freely available. In such a case, AI firms might become less influential, and the technology could lose some of its potency, which might actually be acceptable. Users would still benefit from accessible, functional tools while avoiding another cycle of overhyped promises destined to fail. The concern that AI may hold less value than anticipated could serve as a safeguard against the escalating influence of large tech corporations. If the business model for generative AI is found to be unsustainable, it would be fitting for such an empire to collapse under the weight of its financial statements.