DDR4 has been the workhorse memory standard for a decade. It shipped in Intel Skylake platforms in 2015, became the default for Ryzen in 2017, and still powers the majority of installed business PCs, servers, and industrial computers in service today. Through 2024 and 2025 it remained the cheapest, most available memory on the market.
That is ending. The three largest DRAM manufacturers — Samsung, SK hynix, and Micron — are phasing out DDR4 production through late 2025 and into 2026. The reasons are structural, not cyclical, and the implications for resellers serving customers with DDR4-based installed bases are significant. This is a conversation worth having with customers now, while options still exist.
Why DDR4 is being retired
DDR4 is not being retired because it stopped working. It is being retired because wafer capacity that currently makes DDR4 is more profitable making something else.
The three somethings are DDR5, LPDDR5X for mobile devices, and high-bandwidth memory for AI accelerators. Each of those products commands higher margins per wafer than DDR4. Manufacturers are economically incentivized to reallocate their most advanced fabs away from DDR4 and toward products that extract more revenue from the same production capacity.
This is the same dynamic driving the DRAM pricing surge we wrote about elsewhere in this series. The difference is that DDR4, as the older generation, is on the short end of the reallocation. Newer products get priority; DDR4 production runs get shorter and less frequent.
The counterintuitive part: prices go up on the way out
A common misconception is that end-of-life products get cheaper as manufacturers clear inventory. This is true for some consumer product cycles. It is not true for DRAM.
What actually happens is more like this. Production cuts reduce the flow of new DDR4 into the market. Demand, however, does not drop as fast as supply, because the installed base of DDR4-only platforms is enormous. Customers with DDR4 servers, workstations, and PCs still need memory for upgrades, RMAs, and warranty replacements. That demand chases a shrinking pool. Prices go up.
We saw this with DDR3 during its transition, and we saw it with earlier standards going back to DDR2. Toward the end of a standard's commercial life, the memory gets notably more expensive before it becomes unavailable. Customers who stockpile early pay less than customers who wait for spot market scarcity.
Which platforms are affected
DDR4 is the memory standard for a very broad list of platforms. Any customer running the following is on DDR4:
• Intel 6th through 10th generation Core platforms (Skylake through Comet Lake)
• Intel 11th generation Core platforms running non-DDR5 motherboards
• AMD Ryzen 1000 through 5000 series platforms
• Intel Xeon Scalable 1st, 2nd, and 3rd generation servers
• AMD EPYC Naples, Rome, and Milan servers
• The vast majority of business desktops and workstations deployed before 2023
• Industrial PCs, point-of-sale terminals, and embedded systems with long product cycles
Platforms that use DDR5 natively — Intel 12th generation Core and newer, AMD Ryzen 7000 and newer, Intel Xeon 4th generation Scalable and newer, AMD EPYC Genoa and newer — are not affected.
The three conversations to have with customers
Conversation 1: Inventory the installed base
For any customer with significant PC, workstation, or server fleet on the platforms listed above, the first conversation is about knowing what is in service. Many customers do not have a clean inventory of which machines use which memory type. Help them run the discovery. Count the endpoints, the memory configurations, and the typical refresh cycle length. This inventory is the foundation for everything that follows.
Conversation 2: Forecast replacement demand
Every customer has a baseline rate of memory replacement — modules that fail, upgrades driven by performance complaints, warranty swaps. For a fleet of 500 DDR4 workstations, this is typically a handful of modules per month. Over the three to five years it will take to fully refresh onto DDR5, the customer needs enough DDR4 to cover that baseline. Do the math together: average failure rate times fleet size times expected remaining service life equals the stockpile target.
For customers running DDR4 servers, the math is larger. Enterprise servers get RAM upgrades during their service lives — adding capacity to support more virtual machines, more database tenants, more users. A customer expecting to expand a DDR4 server over the next two years needs to buy that expansion memory now, while it is available at reasonable prices, rather than wait and hope.
Conversation 3: Refresh planning, not just stockpiling
Stockpiling covers short-term risk. For long-term platform health, the customer needs a refresh roadmap. DDR4 platforms installed in 2020 are already five years old; they will likely be refreshed before DDR4 becomes completely unavailable. DDR4 platforms installed in 2023 have a longer runway, but the question of when to refresh them is now a supply question as well as a performance or budget question.
Help customers think about this in terms of platform risk, not just memory risk. If a critical workload runs on a five-year-old Xeon server with DDR4, the question is not just whether the RAM will be available for that server's remaining life. The question is whether the whole server should be refreshed onto a DDR5 platform before DDR4 scarcity becomes a forcing function.
What BCT recommends for the transition
For resellers with DDR4-heavy customer bases, the practical guidance has three parts.
First, help customers buy replacement and expansion DDR4 inventory now, at Q2 2026 prices, rather than wait for spot market conditions later in the year. Inventory is cheaper to hold than it is to scramble for during an outage.
Second, for any customer still deploying new DDR4 systems, ask whether the deployment could shift to DDR5 instead. Some IT teams default to the last generation they are comfortable with. A quick platform evaluation may reveal that DDR5 is the better choice for a new purchase, even if it is slightly more expensive per module, because the long-term availability picture is dramatically better.
Third, treat DDR4 inventory as a managed declining asset rather than a commodity in unlimited supply. Lead times are already longer than they were six months ago. Product line simplifications mean some capacities and speed bins are already unavailable from some manufacturers. The market is not broken — but it is thinning — and resellers who help customers plan for that thinning will win the business that resellers who treat DDR4 as business-as-usual will lose.
Bull Creek Technologies has inventory to continue to offer DDR4 across the full range of speeds, capacities, and form factors, including ECC modules for server applications, registered DIMMs for enterprise workloads, and TAA-compliant UDIMM and SODIMM for federal and regulated-industry customers. For specific part-number availability, lead times, and pricing, contact us at sales@bullcreektech.com