Reply to thread

Message: [QUOTE="Queen, post: 90030, member: 27"] [SIZE=17px][COLOR=rgb(15, 15, 15)]NAND manufacturers cannot erase the current storage shortage by simply turning up production.[/COLOR][/SIZE] Fresh capacity takes years to build, equip, qualify, and stabilise, while AI storage demand is consuming available output now. That gap leaves SSD makers fighting over supply long before any new cleanroom can deliver useful volumes. [HEADING=2]New NAND fab capacity arrives painfully slowly[/HEADING] A modern NAND fab is not an empty building waiting for machines. It needs specialist construction, contamination control, power, water, chemical systems, expensive tools, and a production flow that can repeat microscopic processes without wrecking yields. The NAND fab construction timeline is therefore measured in years rather than quarters. Even aggressive schedules can slip when permits, utilities, tool deliveries, or installation work fail to line up. Equipment does not become productive the day it reaches the cleanroom. Engineers must connect it to the wider line, calibrate each process, run test wafers, inspect defects, and prove that one stage does not destabilise another. NAND equipment lead times also matter during a broad memory boom. Suppliers compete for deposition, etching, inspection, and metrology systems at the same moment, so money alone cannot guarantee immediate delivery. Micron said in June 2026 that NAND supply and demand could remain tight beyond 2027. It also expected industry NAND bit shipments to grow by about 20 percent during 2026, which sounds large until demand is growing faster. That pressure becomes harsher when [B][URL='https://goldmidi.com/community/threads/samsung-supplies-nvidia-nand-squeezing-ssd-supply.75464/']Nvidia's storage appetite drains Samsung NAND[/URL][/B] while competing fabs cannot replace those bits on short notice. A NAND supply recovery needs meaningful output, not a ceremonial first wafer. Early production is usually small, expensive, closely monitored, and reserved for qualification before customers trust it inside high-volume SSDs. [HEADING=2]Higher-layer 3D NAND carries a yield penalty[/HEADING] Manufacturers can create more bits without opening an entirely new fab by moving to denser 3D NAND. That route is faster than greenfield expansion, but it is nowhere near effortless. Higher-layer 3D NAND adds process complexity because deep vertical channels must remain consistent across a taller stack. More deposition, etching, bonding, and inspection steps create more chances for defects to damage usable output. A NAND technology transition can temporarily reduce wafer productivity while equipment is modified and production recipes change. The factory may produce denser chips later, yet supply can feel tighter during the conversion. Yield is the ugly part people skip. A design can look brilliant on paper, but poor early yields mean too many dies fail testing and the promised bit growth never reaches paying customers. QLC NAND bit growth offers another route because four bits are stored in each cell. The extra density helps massive enterprise drives, although tighter voltage margins make endurance, error correction, qualification, and controller tuning more demanding. The 3D NAND production ramp also depends on customer approval. Enterprise buyers test reliability, firmware behaviour, latency consistency, power loss protection, and endurance before placing a new generation into serious infrastructure. This is [B]the painfully slow NAND capacity rebuild[/B]. Layer counts and denser cells raise long-term output, but every transition must survive manufacturing reality before it can improve consumer SSD availability. [HEADING=2]Supplier discipline keeps the shortage alive[/HEADING] The market is not only constrained by engineering. NAND producers remember the brutal oversupply cycle that crushed prices, weakened margins, and forced production cuts earlier in the decade. That history makes suppliers cautious about NAND capital spending. A company can spend billions expanding output, then watch demand soften just as the new line becomes productive. Industry estimates published in early 2026 pointed to restrained NAND wafer output from major Korean suppliers even as AI demand strengthened. Denser QLC production could still lift bit output, but raw wafer starts were not racing upward. Profit also shapes where scarce investment goes. Samsung, SK hynix, and Micron sell several memory products, and high-margin AI memory can win funding, engineering attention, and cleanroom resources over ordinary flash. NAND cleanroom capacity is especially stubborn because existing fabs cannot be rearranged without consequences. Installing new tools may interrupt running lines, while converting space can remove current output before replacement capacity is stable. Enterprise SSD demand gives suppliers another reason to avoid a reckless production surge. Large customers sign longer agreements, buy enormous capacities, and tolerate prices that would send a consumer straight out of the checkout page. Consumer SSD supply will improve only after new capacity, denser processes, healthy yields, and supplier confidence arrive together. Until then, slower price increases can soften the pain, but they cannot manufacture the missing flash. [/QUOTE]

Name