Morning brief · 8:40 AM ET. Afternoon updates post on their own page when the news warrants.

Morning brief 8:40 AM ET

Memory Report — 2026-07-08

Samsung fired the loudest shot of the day, moving its highest-end AI storage into volume and reminding the market that the memory shortage is a full-stack story, not just an HBM one.

The company announced the start of mass production for the PM1763, its newest enterprise SSD designed for AI data centers, built on the PCIe 6.0 interface, which doubles data transfer bandwidth compared with PCIe 5.0. The drive uses Samsung's ninth-generation V-NAND and a new 4nm controller, and is available in 4TB, 8TB, and 16TB capacities, with the 16TB configuration delivering sequential read and write speeds of up to 28,400MB/s and 21,900MB/s, more than 2 times the performance of its predecessor. It slots directly into Nvidia's roadmap: at GTC 2026 in March, Samsung disclosed it would provide HBM4 memory, SOCAMM2 modules and the PM1763 SSD for Vera Rubin, stacking storage on top of its memory position. The demand backdrop is stark. Booming demand from cloud service providers and growing adoption of AI agent services pushed enterprise SSD revenue to a record $18.46 billion in the first quarter, up 86.1% from the previous quarter, and Samsung captured 35% of the enterprise SSD marketplace during Q1, ahead of SK Hynix, Micron, Kioxia, and SanDisk. The earnings echo it: Samsung reported preliminary operating profit of 89.4 trillion won ($58.4 billion) for the April-June quarter, up from 57.2 trillion won in the previous quarter and 4.68 trillion won a year earlier.

On capacity, Micron put shovels in the ground. Over the weekend it broke ground on the expansion of its factory in Hiroshima Prefecture, a ¥1.5 trillion ($9.3 billion) undertaking to produce advanced memory chips, with shipments expected to begin in the summer of 2028, supported by up to ¥500 billion in Japanese government subsidies, and new lines targeting next-generation products like HBM4E for customers including Nvidia. The timing tells the real story of this cycle: a cleanroom started in 2026 does not relieve memory buyers in 2026, and with equipment move-in planned for late 2028, the expansion is aimed at the next phase of AI accelerator demand, not the current shortage.

Pricing stayed firm even as the pace moderated. TrendForce reports further rising contract prices for DRAM and NAND in Q3 2026, with conventional DRAM contract prices expected to rise 13 to 18 percent quarter over quarter and NAND flash 10 to 15 percent. The firm also raised its Q3 PC DRAM forecast to 15%-20%, up from 8%-13%, with Q4 now projected at 3%-8% growth and server DRAM expected to rise 13%-18% on strong demand and tight supply.

The one genuinely red mark is in the equity tape, not the fundamentals. Despite the rout, the Kospi remains the best-performing major index globally this year, up 76% year-to-date, largely on the AI-linked chip rally. Recent sessions have been violent in both directions, driven by index concentration rather than a demand crack: many strategists say foreign selling has less to do with deteriorating fundamentals and more to do with the market's own success. The next test comes fast, with SK Hynix planning to raise around $29 billion on the Nasdaq via ADRs, expected to start trading July 10. Watch that debut and Q3 contract settlements: if pricing power holds through the noise, the structural case built on tight supply and multi-year AI demand stays firmly intact.

July 8th, 2026

Memory Report — 2026-07-08

Sources