Despite DDR5 being the mainstream high-performance standard in today’s time, the exponential gain in price due to the AI boom has made it into unaffordable digital gold bars. While it is true that technological advancement doesn’t wait for anyone, the timing couldn’t be worse, as DDR6 is expected to launch into the market in the next few years.
The “Big 3” Samsung, SK Hynix, and Micron are all moving aggressively into DDR6 development, because the demand coming from AI is only getting more intense, thanks to faster speeds, bigger capacities, and better efficiency. Last year, JEDEC introduced LPDDR6 as the next step forward, improving on LPDDR5 in both performance and power efficiency, and LPDDR memory has basically become the go-to for AI workloads thanks to how well it balances speed and power consumption. That’s also why solutions like SOCAMM2 with LPDDR5/X are starting to show up as a standard approach in AI servers.
Back to DDR6. While official products are still getting prepared, the surrounding ecosystem and supply chain have shown increased activity, with substrate manufacturers being pulled in early. But at the same time, supplies are so constrained right now that it will not solve the current demand as it is, with DDR5 expected to power 90% of servers from 80%+.
If everything stays on track, DDR6 should start showing up in AI data centers around 2028 to 2029, and then trickle down to consumer platforms a bit later once those enterprise demands are satisfied (Emphasizing on the “satisfied”).
