Forecast: spending on generative AI will increase – but mainly on hardware

Global spending on IT products and services related to generative AI will rise to 644 billion US dollars in 2025, according to forecasts by market researchers at Gartner. According to the calculations, this would be an increase of over 75% compared to the previous year. The market researchers assume that generative AI will influence all aspects of the IT spending markets and increasingly become an integral part of business processes and consumer products.

However, this does not mean that AI service subscriptions will suddenly be concluded everywhere. Rather, according to Gartner, the lion's share of 80% of spending in 2025 will be on hardware that manufacturers have upgraded with AI-enabled features. Spending on such servers is expected to climb from 135 billion dollars in the previous year to 180 billion dollars, an increase of 33 percent. And for other AI-enabled devices such as smartphones and PCs, Gartner even expects an increase of almost 100 percent to almost 399 billion US dollars.

Spending on software (around 37 billion dollars) and AI services (around 27 billion dollars) seems almost modest in comparison. However, the expected growth in both segments is explosive, with an increase of 162% for services and almost 94% for software.

Consumers "forced" to use AI hardware

Gartner analyst John-David Lovelock therefore believes that market growth will be strongly influenced by the increasing spread of AI-enabled devices. By 2028, he expects them to account for almost the entire market for consumer devices. "However, consumers are not chasing these features. If manufacturers build AI into consumer devices as a standard feature, consumers will be forced to buy them," Lovelock explains. So far, the marketing efforts for AI computers with Copilot+, which Microsoft and the major notebook manufacturers have been driving since 2024, have not yet been able to generate any major buying impetus.

When it comes to the use of AI in companies, Lovelock assumes that the contradictory trend in the market will continue. On the one hand, high AI error rates and dissatisfaction with the results of proof-of-concept projects are lowering expectations of the technology. "Despite this, base model providers are investing billions annually to improve the size, performance and reliability of GenAI models," says Lovelock.

Lovelock also estimates that companies are putting their ambitious internal AI projects and in-house developments to the test and in some cases even sidelining them. Instead, CIOs would rather opt for commercial off-the-shelf solutions from the various providers, as these presumably offer a more predictable implementation and greater business value.

(axk)