AI Server Boom Could Overtake Smartphone LPDDR Memory Demand by 2027
The race to build more powerful AI systems is beginning to reshape the global memory market. By 2027, demand for LPDDR memory from AI and server platforms is expected to grow so rapidly that it could surpass the smartphone sector, a market that has traditionally been one of the biggest consumers of low-power DRAM.
The biggest driver of this shift is the rise of agentic AI, larger language models, and next-generation AI infrastructure. As AI models become more complex, data centers need more memory capacity, higher bandwidth, and improved power efficiency. LPDDR memory, once mostly associated with smartphones and mobile devices, is now becoming a key part of advanced AI server designs.
According to estimates from Citrini Research, NVIDIA’s upcoming Rubin AI platform alone could require more LPDDR memory in 2027 than the combined demand from Apple and Samsung smartphones. That is a major turning point for the DRAM industry and highlights just how much the AI hardware market is expanding.
NVIDIA Rubin AI racks are projected to consume more than 6,000 million GB of LPDDR memory in 2027. By comparison, Apple’s estimated LPDDR demand for iPhones is expected to reach around 2,966 million GB, while Samsung’s smartphone demand is projected at about 2,724 million GB. Combined, Apple and Samsung are expected to need roughly 5,720 million GB of LPDDR DRAM, meaning NVIDIA Rubin alone could exceed their combined demand by around 6%.
This does not even include other AI platforms that are also expected to require large amounts of memory. AMD’s next-generation MI400 series and AI-focused Helios rack systems are also expected to increase pressure on LPDDR supply. AMD’s Verano CPUs and MI455X accelerators are designed for advanced AI workloads and are expected to support LPDDR5X memory, further adding to demand.
LPDDR memory is becoming attractive for AI servers because it offers a strong balance of capacity, compact design, and lower power consumption. These qualities are especially important in data centers, where energy efficiency and space savings can significantly impact operating costs. As AI racks become denser and more powerful, memory efficiency is becoming just as important as raw compute performance.
Memory manufacturers are already preparing for this shift. Micron has introduced 256 GB LPDDR5X SOCAMM2 memory aimed at future server platforms, while SK Hynix is producing 192 GB LPDDR5X SOCAMM2 memory for NVIDIA’s Vera Rubin platform. Samsung, SK Hynix, and Micron are all increasing their focus on advanced LPDDR standards as they try to keep up with demand from the fast-growing AI ecosystem.
However, current production capacity may not be enough to satisfy both the AI server market and the smartphone industry. New manufacturing facilities are being built to reduce the gap between supply and demand, but expanding memory production takes time and significant investment.
For smartphone brands, this could create serious challenges. If AI companies continue to absorb a growing share of LPDDR supply, phone makers may face higher component costs, tighter availability, and possible production constraints. These pressures could eventually affect device pricing, launch schedules, and overall supply in the consumer electronics market.
The broader technology industry has already experienced the effects of component shortages in recent years, and the rapid growth of AI infrastructure could trigger another wave of supply pressure. As companies compete for high-performance memory, LPDDR DRAM may become one of the most important and contested components in the AI hardware race.
By 2027, the balance of power in the LPDDR memory market could look very different. Smartphones will remain a major source of demand, but AI servers may become the dominant force shaping supply, pricing, and production priorities. With platforms like NVIDIA Rubin and AMD MI400 pushing memory requirements to new levels, the future of LPDDR is quickly moving beyond mobile devices and into the heart of next-generation AI data centers.
