Report says AI revenue has reached tipping point, starting to justify data center investments
A report from research firm Exponential View shows that revenue from AI has reached a critical point, indicating that it may be economically sustainable for technology companies to invest hundreds of billions of dollars in this field. The report shows that global AI sales revenue from hyperscale cloud service providers and emerging cloud service providers reached $25 billion, exceeding the industry’s estimated depreciation costs related to data center and chip investments by $21 billion for the second consecutive quarter. This milestone shows that AI companies are starting to generate enough revenue to cover their capital expenditure costs, but profit margins are still thin. Depreciation charges still consume more than two-thirds of revenue, leaving little buffer to cover other costs such as power, labor and financing. "At present, the economic accounts make sense for the time being," the report said, "but the room for error is very narrow." The findings answer to a central question looming over the AI boom: Whether customer demand is big enough to justify pouring hundreds of billions of dollars into chips and data centers. Meta Platforms Inc. , Alphabet Inc. , Large U.S. technology companies plan to invest as much as $725 billion in capital expenditures this year, a large portion of which will be spent on AI infrastructure. This is one of the largest waves of corporate spending 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 type of capital expenditure investment, you should not expect it to significantly cross this threshold; because if it does, it may mean that you miss some opportunities that you could have seized." The AI craze has largely been measured from the supply side, based on disclosures from listed semiconductor companies such as Nvidia and hyperscale cloud service providers such as Alphabet. The demand side is harder to quantify because many of the most important AI labs, including OpenAI and Anthropic, are still not listed. The data is based on a dataset built by Exponential View, which tracks AI spending by more than 1,000 companies. It uses sources including company documents, executive comments, news reports and cloud service provider disclosures, and adjusts the data to avoid double counting between layers of the AI supply chain. The analysis assumes a six-year depreciation life for information technology (IT) equipment, including graphics processing units (GPUs), the chips used to train and run advanced AI models. Some investors think this assumption is too optimistic because the rapid pace of chip innovation could devalue older hardware in just a few years. However, data in the report shows that the value of older chips has not collapsed. The hourly rental price of NVIDIA H100 chips currently remains at nearly 80% of the level at the time of release. "Even entering its fourth year, it's still completely in demand," Azhar said. He pointed out that the rental price of the chip has increased in the past year as the demand for AI computing power exceeded the supply of Nvidia's new Blackwell chip.