‘Unlimited demand for intelligence’: Why a collapse in AI costs is reshaping labour, power and governance

A steep fall in the cost of artificial intelligence has reignited debate over the future of work, power and governance, after a social media post drew attention to what one user described as a “civilizational-scale” shift in how intelligence is produced and deployed. 

The discussion was sparked by Aakash Gupta, a social media user, who posted on X (formerly Twitter) about the dramatic collapse in AI pricing over the past two years and its implications for the global economy and labour markets. 

“GPT-4 launched at $60 per million output tokens. Today, equivalent capability costs under $1. That’s a 98% price collapse in two years. Demand didn’t fall. It exploded,” Gupta wrote. He pointed out that OpenAI’s annual recurring revenue (ARR) jumped from $1 billion to more than $12 billion even as prices were cut quarter after quarter. 

Gupta framed the shift through the lens of Jevons Paradox, an economic principle that holds efficiency gains often lead to higher overall consumption rather than conservation. Drawing a historical parallel, he wrote: “When coal got cheaper in the 1800s, England didn’t use less coal. They burned 10x more. Intelligence is following the same curve, except the adoption rate is compressing a century of energy economics into 36 months.” 

According to Gupta, the consequences extend far beyond automation. Citing Stanford research, he highlighted a 280-fold reduction in AI compute costs between 2022 and 2024, noting that tasks once costing $1,000 can now be completed for under $4. 

“At that price, companies don’t just automate what humans were doing. They start doing things that were never economically viable at human-labor pricing,” he said, adding that analytical work once requiring a highly paid analyst for a year can now be completed “for $50 in an afternoon.” 

As intelligence becomes abundant and cheap, Gupta argued, it ceases to be the scarce input. “Taste, judgment, and the ability to ask the right question become the bottleneck,” he wrote. “The returns flow to people who can direct intelligence, not people who provide it.” 

Gupta’s post was shared alongside another user’s assertion that “There is unlimited demand for intelligence,” a claim that drew both agreement and concern across the platform. 

One reply warned that the central risk lies not in abundance, but in unchecked scale. “‘Unlimited demand for intelligence’ is true. The dangerous part is what people conclude from it,” the user wrote. “When intelligence gets cheap, the scarce resource is not ‘taste.’ It is stability.” 

The commenter cautioned that cheap AI does not simply replace labour — it scales decisions, often faster than institutions can regulate them. “When coal got cheaper, we did not just burn more coal. We built machines that could burn it faster than we could regulate the consequences. AI is that, but for cognition,” the post said. 

The real bottleneck, the user argued, is the absence of safeguards, citing the need for reliability, auditability, constraint enforcement, provenance, adversarial testing, rollback mechanisms, human legibility and institutional accountability. 

“Otherwise you get a world where every company runs 10,000 autonomous analysts, shipping decisions into production at machine speed, with no coherent oversight,” the user added. “Cheap intelligence is not the endgame. Regulated intelligence is.” 

A third voice questioned whether demand for intelligence would actually keep pace with its rapidly expanding supply, suggesting the trend could deepen inequality instead. 

“I think the current trajectory is to get to the point of almost infinite offer of intelligence,” the user wrote. “Demand will not grow at the same pace as the offer and will likely redefine the labour market.” 

The commenter predicted a more polarised society, where individuals and businesses capable of effectively leveraging AI reap disproportionate gains, while others fall behind. Despite dramatic reductions in unit costs, they argued that “the average user of AI platform is very far from being able to increase their productivity/output in a meaningful way.” 

What is Jevons Paradox? 

Jevons Paradox holds that when a resource becomes cheaper or more efficient to use, total consumption often rises rather than falls. First observed in the 19th century by British economist William Stanley Jevons, the phenomenon emerged when more efficient steam engines led Britain to burn far more coal, not less. Cheaper energy made new applications viable, expanding demand across industry, transport and manufacturing. 

The same logic is now being applied beyond energy, particularly to AI and computing. As AI becomes cheaper, faster and more accessible, it does not merely replace existing tasks — it enables entirely new uses that were previously uneconomic. That, proponents argue, is why falling AI costs may fuel an explosion in demand for intelligence rather than constrain it.