
Gartner Sees Datacenter Hardware Spend Hitting $1.73 Trillion in 2027
Gartner's November forecast puts 2027 datacenter systems spending at $1.73 trillion, up 48 percent, overtaking enterprise software for the first time as AI hardware demand reshapes IT budgets.
- By
- Nathan Brooks
- Filed
- Channel
- AI & Compute
- Read
- 4 min read
Gartner's latest forecast puts global spending on datacenter systems at $1.73 trillion in 2027, up 48 percent year on year — enough to surpass enterprise software, at $1.68 trillion, for the first time. Hardware is back in the driver's seat. It has not sat there since before the cloud ate the datacenter.
The figures come from Gartner's November 2026 IT spending forecast, published just before the US elections, which also carries the research firm's first projections for 2027. Enterprise software still grows — a very normal 14.7 percent next year — but the hardware line simply grows faster. IT services, with infrastructure as a service carved out, will rise a more anemic 4.9 percent, roughly twice the pace of global GDP. That, by the standards of mature markets, is business as usual.
What is new in the forecast?
Gartner has split IaaS — rented cloud compute, storage and networking — out of its sprawling IT services segment and given the category its own line. The change is overdue. Buyers of iron and renters of capacity now appear side by side, and the analyst only published 2026 and 2027 figures for the new category. By working backward from the percentage changes, and stitching in published Gartner IaaS figures from 2015 through 2024, the longer trend becomes visible.
The picture it shows is stark. For years, enterprise software revenues ran two to three times the size of datacenter systems sales. AI systems priced at genuinely absurd levels, plus the upgrades to back-office transactional systems and front-office analytics that feed them, have closed that gap in a single cycle.
How fast is hardware actually growing?
Very, and from a standing start. Gartner expects datacenter systems revenue growth of 73.8 percent in 2026, cooling to a "mere" 55.4 percent in 2027. Between 2013 and 2021 inclusive, the average was 2.3 percent a year — below the 2.7 percent average global GDP growth over the same period. The systems market was beyond mature, sitting in the trunk of the IT economy while software and services drove.
What changed is stinginess. Companies can no longer afford to miss the GenAI wave, so the hardware budget has been unlocked. In that respect the boom resembles the dot-com era, with one difference: the hyperscalers and cloud builders already exist, and they are hogging the GPU and XPU allocations.
The composition of IT spending has shifted accordingly. In 2012, core IT — systems, software and services — accounted for a little more than a third of overall IT spending. By the close of 2027 it will exceed two thirds. The gap between core IT growth and global GDP growth is not the normal-ish 2X but roughly 8X in 2026 and 7.3X in 2027, if Gartner's spending numbers and the World Bank's GDP expectations hold.
We are not living in normal times. Clearly.
How much of the spend is real capacity?
Some of the surge is money buying more machines, and some is money buying the same machines at higher prices. Adjusting datacenter spending figures for currency inflation against the 2021 dollar gives a partial answer. The rest — price inflation from too much demand chasing too little supply — resists precise measurement.
The honest estimate is roughly half and half: about half of this year's growth in datacenter systems spending reflects supply-demand pressure that drives up prices, and half reflects actual increases in delivered supply from chip and system makers. The source calls that an educated guess. It resembles nothing so much as buying plywood and contractor services during the pandemic — when everyone wants the same scarce thing at once, the bill rises faster than the pile.
Is 2027 the peak?
Plenty of prognosticators think so — not because demand fades afterward, but because political resistance to tech titans and their datacenter and electricity appetites is rising just as it grows unlikely that power generation and distribution can ramp fast enough to meet datacenter demand in 2030, and less so in 2028. Forecasters themselves concede the data is wonky and wobbly, and that predictions much beyond twelve to eighteen months drift with it.
Still, buyers plan their own budgets against these numbers, which is why the revisions matter even when they are wrong. Gartner's trajectory says hardware keeps outrunning software through 2027, at growth rates the industry has not printed since the internet was young. The electricity, and the politics, may decide what happens after.
Source: The Next Platform
More from Nathan Brooks
Show full bio
Senior reporter covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
786 articles
Related articles
memory-capacity-ceiling-slows-ai-spending-forecasts-gartner-data-shows-fae40af6
Memory Capacity Ceiling Slows AI Spending Forecasts, Gartner Data Shows
gartner-puts-genai-model-spend-at-28-3-billion-c7818787
Gartner Puts GenAI Model Spend at $28.3 Billion for 2026
datacenters-will-eat-3-3-of-world-power-gartner-says-06f3688c
Datacenters Will Eat 3.3% of World Power by 2030, Gartner Says
ai-data-center-boom-meets-capital-power-and-grid-constraints
AI Data Center Boom Meets Capital, Power and Grid Constraints



