The high demand of artificial intelligence may be the destruction of its profit margins star-news.press/wp

In recent months, Silicon Valley has become fond of 160 -year -old economic theory from the Railway and Coal Age: The Jevons Paradox.

It is originally applied to steam engines and energy consumption, and the concept determines that efficiency gains do not reduce the use of resources. Instead, they are He increases This, by making this resource more useful. Dynamics appeared in the era of railways, thus: the engines become more efficient in energy, which require less coal, but the demand for coal still explodes. Why? Because efficiency gains have pushed a broader adoption of energy -saving engines.

In our time, Jevons Paradox is applied to artificial intelligenceWhich, such as railways and steam engines, requires huge amounts of energy and thus depending on the power supply. However, reinforcements such as Microsoft Satya Nadella CEO of Microsoft Satya Nadella tend to transfer the energy side of the equation, Summoning Jevons paradise If you simply suggest that, taking into account artificial intelligence, the demand is more efficient, and the demand will explode, not contracting – pushing the profits up.

But one Industry researcher He argues that this vision, although it is tempting, is also misleading. In the influencing Postcal PostIndependent analyst Dave Friedman argues that the Javsnis paradox is not a footnote to the economy of artificial intelligence – it is the entire plot, and it indicates dangerous and growing restrictions as the profitability of artificial intelligence companies relates to.

His argument? Yes, artificial intelligence may become more efficient. At the same time, the enlarged demand devoured any efficient gains. And

The problem is the most obvious at the small end of the market funnel spectrum, among startups. As Friedman IndicateEmerging companies of artificial intelligence Photograph For investors as programs similar to programs, known as Saas (Software-AA-ARVICE), a famous company of famous margin and golden rush, a modern memory that does not exceed the VC Hearts race.

With the Saas Business model, it does not cost companies much more to serve additional customers, and therefore the subscription fees for each additional customer decreases in one way or another to a minimum. The best saas plays may be margins up to 80 or 90 %, which puts them among the best companies in the world in this regard and attract capital like flies to honey.

But startups artificial intelligence Just resemble software companiesFriedman argues, due to the paradox of Givelon. With artificial intelligence companies – unlike the Saas companies – the energy costs and related costs are not determined. Therefore, additional customer service and a broader accreditation does not make Amnesty International cheaper. Expenditure scale with use.

“This dynamic explains the reason for the emergence of many emerging companies of artificial intelligence on paper, but it may never reach healthy profit margins,” said Friedman, a startup and former Citigroup and Bloomberg. “If your primary product is more expensive to provide more things that are used, this is a structural problem, not a temporary defect.”

From Friedman’s point of view, Silicon Valley controls physics. In his participation, he wrote, in reference to the way some companies depict that the costs of reasoning are OP-EX, in reference to the way some companies depict the continuous computing needs of artificial intelligence as fixed expenses (or at least almost installed) under GAAP accounting. In fact, he says, these costs rise with use, so it will be better to understand it as gears, or the cost of sold goods, which reflect more accurately the costs of use.

In fact, in some other energy -intensive industries, energy use and cost calculation are calculated as varieties. For example, the energy used is delicious one ton of aluminum directly to this type It included in Cogs. The same applies to some technical companies. Includes AWS Server and cooling electricity as anxiety When you are linked to the connection of customer work burdens.

It may seem to be a simple accounting difference, but in practice, it may be the difference between an accurate image of the basic business economics – an inaccurate image.

Worse, Friedman said, The wrong belief is widespreadPartially due to the distortion factor: calculating the credits. Many LLM startups work on free or dramatically reduced infrastructure such as AWS, Google and Microsoft. This can already hide the cost of customer service and make P&Ls look artificially healthy.

Historically, this is not new-from communications to the cloud to energy, and investors may exaggerate and repeated early margins in heavy infrastructure. What makes Amnesty International, says Friedman, is the speed. “A penetration mode can think about your entire stability overnight,” Quartz told Quartz. “And the open source community comes out of these breaches faster than ever.”

The dynamic is a special punishment for startups. Giants such as Google and Microsoft possess their databases, negotiate electricity at industrial prices, and have eliminated contracts to build infrastructure that occupies artificial intelligence. In contrast, startups stumble in renting such resources, and often from the same positions they hope to disable. This makes them very sensitive to the cost structure.

“Moreover, they do not have a financial pillow to accommodate the margin that strikes the way in which large technology can master,” said Friedman. “Even if two companies are gaming the same form, starting starts pays much more, and feels it more severely.”

This does not mean that we are ready to see the collective failure of emerging companies in artificial intelligence and LLM. But it can mean reset. From Friedman’s point of view, many emerging companies from artificial intelligence will be forced into uncomfortable decisions: trimming the most accumulation combines, high prices, or stability in permanently low margin operations.

“They will survive, but they will not look like classic Saas companies with 80-90 % of margins. It will look more like infrastructure companies, as profits are thinner and growth takes longer,” said Friedman.

In other words, the Jevons paradise of our time may mean more efficiency and more demand – but for startups, at least, not necessarily more money.