“Memory Flowing to AI Sends General-Purpose Server Prices Up 4x”... Enterprise IT Faces a Lean Period
IT DAILY · · 1 views
✦ AI Summary
Gartner said that AI data center investment has reduced supply of memory and SSDs for general-purpose servers and storage, driving prices sharply higher.
General server prices this year are 2 to 4 times higher than in 2025, storage prices are up to 2 times higher, and a high-end x86 server with 1.5TB of memory and four 3TB NVMe SSDs has risen 3.5 times from the fourth quarter of 2025.
Gartner said this high-cost structure is likely to last at least through 2028 and recommended normalizing IT budgets, accelerating purchase approvals, avoiding long-term contracts, and optimizing architecture.
AI-focused investment is driving up prices for general enterprise IT infrastructure. As funding and demand concentrate in AI data centers amid large-scale investments by global hyperscalers in AI data centers, competition for GPU and HBM supply has intensified, and memory shortages for general-purpose servers and storage have also worsened.
As a result, prices for general-purpose servers and storage for enterprise computer rooms have surged. Gartner released a report titled "Server and Storage Costs in 2027 Amid the AI Boom," saying that general server prices this year are 2 to 4 times higher than in 2025, while storage prices have risen to as much as 2 times this year.
Gartner said prices for a high-end x86 server equipped with 1.5TB of memory and four 3TB NVMe SSDs jumped 3.5 times from the fourth quarter of 2025. It explained that this was the result of the ripple effects of expanded AI data center investment, which have put pressure not only on GPU and HBM supply but also on memory supply for general servers and storage.
Gartner forecast that this high-cost structure will continue at least through 2028. It said prices may gradually decline after that, but are unlikely to fully return to 2025 levels.
The fundamental cause of the price spike is being attributed to a large-scale reallocation of the memory supply chain. Major manufacturers such as Samsung Electronics, SK hynix, Micron, and Kioxia have prioritized DRAM and NAND flash production capacity for HBM for GPU and hyperscaler AI infrastructure, sharply reducing supply of memory modules (DIMM) for general servers and also cutting supply of enterprise SSDs.
As a result of these changes, memory is expected to account for a much larger share of total costs in non-AI general servers. Gartner put the memory share of total non-AI general server costs at 45% in 2025, and projected that it will reach two-thirds by 2030.
On top of that, as supply chain bottlenecks and longer delivery delays persist, the timing for new Fab capacity from memory makers was cited as the latter half of 2027. Server and storage vendors are likely to delay price cuts to defend margins, so the price stabilization felt by buyers is bound to be slow.
Lead times for custom server and storage orders have stretched from the existing 4 to 6 weeks to 3 to 5 months. Vendors are steering purchases toward specific specifications that are available, excluding ultra-high-capacity memory and SSD options. The longer procurement period for customized on-premises equipment and the shrinking range of available specifications have increased cost burdens.
This hardware cost shock in on-premises environments has spread to the public cloud (IaaS) market. In the wake of higher hardware procurement costs, IaaS fees were raised earlier this year, with Akamai, OVHcloud, and Tencent Cloud each increasing prices by 5% to 15%. Hetzner Cloud's price hikes reached as much as 100%.
Amazon Web Services (AWS) and Microsoft Azure kept list prices unchanged. However, AWS and Azure imposed capacity limits in some regions, including blocking new deployments. As a result, customers were steered to other regions or unwanted VM choices, effectively creating a price increase.
Gartner said that, to respond to this situation, companies need to overhaul their budgeting, negotiations, and infrastructure operating strategies.
Gartner stressed the need to normalize IT budgets for 2027 to 2028 based on the current high-price environment. It also said purchase approval procedures need to be accelerated in light of lead times. For negotiations, it said companies should use their own "appropriate cost model," explaining that the model is based on 2025 purchase prices with memory and SSD increases of 4 to 5 times reflected. It added that the purpose of the model is to verify vendors' excessive margin demands.
Gartner said server prices are expected to fall in 2028 and recommended delaying large-scale purchases until then. It also said long-term contracts that lock companies into unfavorable prices beyond 2027 should be avoided. It further stressed that architecture optimization to maximize existing infrastructure resources is essential. It recommended extending server replacement cycles to 6 to 8 years and raising the memory oversubscription ratio in VM environments.
For storage strategy, it proposed Tiering that combines low-cost HDDs and ultra-high-capacity SSDs (30 to 120TB). It also said storage should apply data reduction technologies of at least 4:1. Gartner said companies should actively consider STaaS to avoid the initial Capex shock, and emphasized that cloud storage migration should also be actively considered.
Source: IT DAILY · Kwon Young-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=242220
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.