According to a report from research firm Exponential View, revenue from artificial intelligence has reached a critical threshold, indicating that the hundreds of billions of dollars being invested in the sector by tech companies may be economically sustainable.
The report shows that global AI sales revenue for hyperscale and emerging cloud service providers reached $25 billion, exceeding the estimated depreciation costs of $21 billion associated with the industry's data center and chip investments for the second consecutive quarter. This milestone suggests that AI companies' revenues are beginning to cover their capital expenditure costs, though profit margins remain slim. Depreciation expenses still consume over two-thirds of the revenue, leaving a very small buffer to cover other costs such as power, labor, and financing.
"For now, the economics are temporarily adding up," the report states, "but the margin for error is narrow."
These findings address a central question looming over the AI boom: whether customer demand is substantial enough to justify the hundreds of billions poured into chips and data centers. Major U.S. tech companies like Meta Platforms Inc., Alphabet Inc., Microsoft, and Amazon plan capital expenditures of up to $725 billion this year, a significant portion of which is earmarked for AI infrastructure, representing one of the largest corporate spending waves in history.
"It has just crossed the depreciation threshold, and roughly speaking, the situation is gradually improving," Azeem Azhar, founder of Exponential View and an investor in dozens of startups, told the media. "At this stage of any capital expenditure investment, you shouldn't expect it to significantly surpass that threshold; if it did, that might actually mean you missed some opportunities you could have captured."
The AI boom has largely been measured from the supply side, based on disclosures from listed semiconductor companies like NVIDIA and hyperscale cloud providers like Alphabet. The demand side is harder to quantify, as many of the most important AI labs, including OpenAI and Anthropic, remain private.
This data is based on a dataset built by Exponential View that tracks AI spending across more than 1,000 companies. Its sources include company filings, executive commentary, news reports, and cloud provider disclosures, with adjustments made to avoid double-counting across different layers of the AI supply chain.
The analysis assumes a six-year depreciation period for information technology (IT) equipment, including graphics processing units (GPUs), the chips used to train and run advanced AI models. Some investors consider this assumption overly optimistic, given the rapid pace of chip innovation that could render older hardware obsolete within just a few years.
However, data in the report suggests the value of older chips has not collapsed. The hourly rental price for NVIDIA's H100 chip remains at nearly 80% of its initial launch level. "Even entering its fourth year, it is still fully in demand," Azhar noted, pointing out that the chip's rental price has increased over the past year as demand for AI computing power has outpaced the supply of NVIDIA's new Blackwell chips.
Comments