Counterpoint warns that DDR5 RDIMM costs may surge 100% amid manufacturers’ pivot to AI chips and Nvidia’s memory-intensive AI server platforms, leaving enterprises with limited procurement leverage.
Server memory prices could double by the end of 2026, driven by manufacturers’ pivot to AI-focused production and Nvidia’s shift to smartphone-style memory that is straining supply further.
DRAM prices have already risen approximately 50% year to date and are expected to climb another 30% in Q4 2025, followed by an additional 20% in early 2026, Counterpoint Research said in its report. The firm projected that DDR5 64GB RDIMM modules, widely used in enterprise data centers, could cost twice as much by the end of 2026 as they did in early 2025.
At the same time, the firm also expected DRAM output to grow by more than 20% in 2026 as manufacturers ramp production, but warned this may not be sufficient to offset AI-driven demand.
Enterprise buyers can already feel the impact. Procurement teams are facing price increases for servers, PCs, and smartphones, with steeper hikes expected for systems built with next-generation memory configurations starting next year, the firm added in its report.
The firm’s analysis highlighted that Nvidia’s LPDDR-based AI server platforms are creating unprecedented demand in that memory segment. According to Nvidia’s specifications, the Grace CPU Superchip uses up to 960GB of LPDDR5X memory, compared to 16GB in a premium smartphone. The requirements are expected to grow further with the upcoming Vera CPUs.
“The bigger risk on the horizon is with advanced memory, as Nvidia’s recent pivot to LPDDR means they’re a customer on the scale of a major smartphone maker — a seismic shift for the supply chain which can’t easily absorb this scale of demand,” MS Hwang, research director at Counterpoint Research, said in the report.
Limited options for enterprise buyers
As supply tightens, most enterprises face limited leverage in selecting suppliers.
“Enterprise will have less control over what memory supplier they can choose unless you are a hyperscaler or tier-2 AI datacenter scale enterprise,” Neil Shah, VP for research and partner at Counterpoint Research, told NetworkWorld. “For most enterprises investing in AI infrastructure, they will rely on vendors such as Dell, Lenovo, HPE, Supermicro, and others on their judgment to select the best memory supplier.”
Shah advised enterprises with control over their bill of materials to negotiate and lock in supply and costs in advance. “In most cases for long-tail enterprises, smaller buyers without volume leverage, they will have little control as demand outstrips supply, so the prudent thing would be to spread out the rollout over time to average out the cost spikes,” he said.
Legacy shortage opens door for Chinese suppliers
The current pricing pressure has its roots in production decisions made months ago. According to Counterpoint, the supply crunch originated at the low end of the market as Samsung, SK Hynix, and Micron redirected production toward high-bandwidth memory for AI accelerators, which commands higher margins but consumes three times the wafer capacity of standard DRAM.
That shift created an unusual price inversion: DDR4 used in budget devices now trades at approximately $2.10 per gigabit, while server-grade DDR5 sells for around $1.50 per gigabit, according to the firm.
This tightness is creating an opportunity for China’s CXMT, noted Shah. “DDR4 is being used in low- to mid-tier smart devices and considering bigger vendors such as Samsung and SK Hynix planned to ramp down DDR4 capacity, CXMT could gain advantage and balance the supply versus demand dynamics moving into the second half of next year,” Shah said.
Enterprise adoption of Chinese suppliers will depend on regional considerations, Shah noted, with some organizations being comfortable with CXMT and others requiring non-Chinese sources depending on compliance requirements.
Error correction approach changes
Beyond supplier considerations, enterprises also face technical differences with LPDDR-based systems. Nvidia’s shift to LPDDR changed how error correction works in server environments, though Shah clarified this does not eliminate protection.
“Even though AI servers are expanding the use of LPDDR instead of DDR, most memory chips do have on-die ECC,” he said. “Memory and CPU will continue to work in tandem to actively check before the data is transmitted out and after during transmission, respectively.”
However, enterprises will need to adapt validation processes for error checking, scrubbing, and testing component health, including thermals and bit error detection, Shah added.
Major vendors are already raising prices
These supply and technical challenges come amid an already tightening market. Samsung has already raised prices on 32GB DDR5 modules to $239 from $149 in September, a 60% increase. SK Hynix reported during its October earnings call that its HBM, DRAM, and NAND capacity was sold out through 2026, while Micron raised prices 20-30% and stopped quoting some products entirely.
Samsung, SK Hynix, Micron, and Nvidia did not immediately respond to requests for comment.
