A $961 billion increase that is mostly a hardware bill
Gartner published its latest worldwide IT spending forecast on October 21, projecting that global IT spending will hit $7.35 trillion in 2027, up 15 percent from $6.39 trillion in 2026. That alone is a milestone: it is the first time the firm has forecast IT spending crossing $7 trillion. But the number that should get a CTO's attention is the composition of that growth. Of the roughly $961 billion increase, data center systems alone account for about $487 billion of it, or just over half.
That is not a rounding error. Software, the next-largest contributor, adds $215 billion. Services adds $76 billion. Infrastructure as a service adds $75 billion. Devices add $69 billion. Communications services add $39 billion. Stack all five of those together and they still do not match what data center hardware alone is projected to add. If you run a budget that includes any meaningful on-premises or colocated footprint, or if your cloud bill passes through hardware refresh costs in any form, this forecast is telling you where the pressure is concentrated.
Memory prices, not AI adoption, are driving the number
Gartner's explanation for the spike is not demand growth in the way most CIOs would assume. It is price. John-David Lovelock, the firm's distinguished VP analyst, points to what Gartner is calling a memflation effect: memory prices have more than doubled in the past 12 months, and Gartner expects that pressure to persist through 2027 before hardware pricing begins to stabilize. Data center systems grew 73.8 percent in 2026 and are forecast to grow another 55.5 percent in 2027, almost entirely on the back of memory and AI-optimized server pricing rather than unit volume.
That distinction matters for how you read every other AI infrastructure headline this quarter. A forecast built on rising average selling prices is not the same thing as a forecast built on rising demand. Gartner's own commentary makes this explicit: higher prices are supporting the spending totals even as affordability pressure is expected to suppress end-user demand. In plain terms, enterprises are not necessarily buying more data center capacity next year. Many of them will pay more for the same capacity, and some will buy less of it than they planned.
Every other category grows in single digits by comparison
Line up the segments and the imbalance is stark. Software grows 14.6 percent in 2027. Services grow 4.9 percent. Infrastructure as a service grows 25.7 percent, down from 31.3 percent growth in 2026. Devices grow 8.1 percent. Communications services grow 2.9 percent. Data center systems grow 55.5 percent, on top of 73.8 percent growth the year before. No other category comes close, and none of them is accelerating the way data center hardware is. IaaS, the category most directly tied to public cloud consumption, is actually decelerating even as the underlying hardware beneath it gets more expensive, which is its own signal about where margin pressure is heading next for the providers running that infrastructure.
That asymmetry is the real story for anyone building a 2027 technology budget. If your organization's IT spend mix looks anything like the market average, data center infrastructure costs are going to grow faster than your software licensing, your managed services contracts, and your device refresh cycles combined, and by a wide margin. Finance teams that built 2027 plans off a flat or modestly rising hardware line are working from an assumption Gartner's own numbers do not support, and the gap will show up first in server, storage, and GPU-adjacent line items. The usual cost levers will not help much here either: renegotiating a software license or consolidating a services vendor can trim a budget line growing in the single digits, but it does little against a hardware category growing north of 50 percent a year on pricing alone, because the lever that matters there, memory and component cost, sits entirely outside the enterprise buyer's control.
AI infrastructure is becoming an asset class, which changes who controls pricing
Lovelock frames the shift in language that is unusual for an IT spending forecast: AI infrastructure, he says, is emerging as both a technology platform and an investable asset class. That framing explains a lot of what has shown up elsewhere this quarter, from hyperscaler capex guidance to colocation debt deals to the data center financing structures now routinely rated by S&P and Moody's. When infrastructure becomes an asset class, the people setting prices are no longer only the hardware vendors and the cloud providers. They include the debt and equity markets financing the build-out, and those markets price in scarcity, not customer budgets.
Gartner also flags surging electricity demand from AI and data centers as a separate driver pushing enterprises toward direct energy investment, a trend that compounds the hardware pricing problem rather than offsetting it. Lovelock's framing for 2027 growth is explicit: it depends on the ability to finance, build, and monetize AI infrastructure at an unprecedented scale. That is a statement about capital markets, not about whether your workloads actually need more compute.
What this means for the 2027 roadmap
The practical implication is straightforward even if the macro picture is not. Do not assume cloud unit costs will keep falling the way they did for most of the past decade. Memory-driven hardware inflation gives every major provider cover to hold or raise prices through at least 2027, and the forecast suggests they will have the pricing power to do it regardless of whether your actual usage grows. If you have renewal leverage with a hyperscaler or colocation provider, use it now, before the next pricing cycle locks in at this quarter's inflated baseline.
It is also worth separating this forecast from the demand story you are hearing elsewhere. Rising average selling prices can make a market look hotter than it is. Gartner's own affordability warning should be read as a caution against over-provisioning on the assumption that everyone else is scaling up. Some of your peers are paying more for the same footprint, not expanding it, and your 2027 capacity planning should start from that distinction rather than from the headline growth number.
