Key insights
- Mizuho analyst Jordan Klein is bullish on DRAM and memory stocks due to the AI CPU acceleration narrative. He highlights tightening supply conditions and rising prices in the memory market. Micron (MU) and SanDisk (SNDK) are noted as undervalued. Increased CPU attach rates to GPUs in AI applications, as emphasized by Intel, further support the demand for memory. This suggests a positive outlook for memory stocks and a potential boost for the semiconductor sector.
Investing.com -- A surge in artificial intelligence (AI) investment is reshaping the semiconductor landscape, with investors increasingly focused on the interplay between processors and memory chips that power data-heavy workloads. Mizuho TMT specialist Jordan Klein has named several memory stocks investors looking to benefit from the AI CPU trade should be buying up, many of which are trading at attractive levels.
“I could not be more bullish on DRAM and Memory overall off this Ai CPU acceleration narrative/chase,” Klein said.
The call comes as industry data points to tightening supply conditions. With supply falling short and no new supply expected until 2H27, prices are expected to go up. HBM (high bandwidth memory), a specialized form of DRAM used in AI accelerators, has a 3-1 trade ratio.
“You cannot operate a cpu without ram memory. Do not forget that fact. They go hand in hand. So a lot more server (or even client CPUs) means a lot more DRAM over time,” he added.
Among NAND stocks, he notes that Micron Technology (NASDAQ: MU), Samsung (KRX: 005930), Hynix (KRX: 000660) and SanDisk Corp (NASDAQ: SNDK) are “way cheaper” than Advanced Micro Devices (NASDAQ: AMD), Arm Holdings (NASDAQ: ARM) and Intel Corp (NASDAQ: INTC).
MU is “crazy cheap“ at 3-4x buyside EPS, and SNDK also “still feels good” ahead of earnings with SSD demand rising the possibility that management will guide for $25-30 per quarter on EPS with an 80%+ gross margin.
“Feel like could get bought even after this massive YTD rally and being pretty owned across the momentum investor crowd,” Klein said, adding that he doesn’t think many hedge funds and active fund managers are Overweight or long SNDK, given the high volume risk in the stock.
One of the key takeaways from INTC’s conference call is the focus on a much higher CPU attach rate to GPU in agentic AI applications, where AI systems autonomously plan and execute tasks, which could move from 1:8 to 1:2 and even 1:1.
“INTC mgmt. communicated on the call that with new Ai coded agentic applications, the need for CPU to manage and orchestrate the tasks is much higher vs training or creative content Ai applications that are way more GPU or ASIC intensive. You need a lot of security and planning,” the analyst stated.
According to Klein, AMD is best positioned with CPU, FPGA, and GPU, and less capacity constrained than INTC as they use TSM. He also expects Nvidia (NASDAQ: NVDA) to sell “a lot of” ARM based CPUs for its new AI workload.