Dive Brief:
- Global spending on AI hardware nearly doubled in the first half of 2024, reaching $47.4 billion, IDC said in a Tuesday report. The technology market intelligence firm expects AI infrastructure investments to surpass $200 billion annually in 2028.
- Spending on AI servers grew 105%, accounting for 95% of global AI infrastructure spending during the first half of 2024, the analysis revealed. AI storage hardware investments increased by just 18% during the six-month period.
- The spending spike marked an inflection point as investments in AI compute and storage capabilities are expected to surge past $120 billion in 2024, more than doubling year over year. The market sustained double-digit annual growth since 2019 and entered an accelerated growth pattern in 2023, said Lidice Fernandez, IDC group VP, Worldwide Enterprise Infrastructure Trackers.
Dive Insight:
Big tech took a commanding lead in high-capacity processing power last year. Cloud and digital service providers accounted for nearly three-quarters of spending on AI servers for the period IDC analyzed. Traditional enterprises, in contrast, were more cautious and constrained in their hardware investments.
“Component availability was horrific over the last two years,” Fernandez said. “If you were a traditional enterprise, not a large data center, your wait time could be 36 weeks to get your hands on a server with an accelerator.”
Tech talent availability and cost of ownership slowed enterprise AI deployment plans, too.
“It’s a combination of not being sure which projects will benefit from moving into an AI environment and not having the skills to build and maintain the infrastructure,” Fernandez said. “You can end up spending $250,000 for one or two of these servers and you’re going to need infrastructure skills to justify the total cost of ownership.”
The largest hyperscalers — AWS, Microsoft and Google — were unbridled in their investments. Each of the three providers poured tens of billions of dollars into AI infrastructure buildouts last year and plan to maintain high levels of spend through the first half of 2025.
Cloud’s massive appetite for AI hardware also left its mark on the semiconductor market. Hyperscaler spend on AI chips and processors nearly doubled to $112 billion last year, as the global market surpassed $626 billion, according to Gartner research.
AI storage hardware followed a different path. Cloud providers accounted for just 40% of global spend in the category, largely due to the cost differential separating storage from compute, according to IDC.
“The trend for storage is to become less expensive as time goes by,” Fernandez said. “Even though capacity continues to increase, prices keep dropping.”
As AI becomes embedded in cloud and enterprise business processes, hardware to support the technology will continue to spread. Fernandez said AI servers and storage comprised roughly one-third of total infrastructure spend in 2023. IDC expects it to amount to half by 2028.
