Level 1 — Absolute Beginner

There is a company in the United States called Seagate. It makes special parts that store information for computers.

On May 19, 2026, the boss of Seagate spoke to people who buy and sell stocks. His name is Dave Mosley. He said the company cannot make enough parts because of computers that use AI.

After he spoke, the prices of stocks in some big computer-parts companies went down. Seagate, Western Digital, Micron, and SK hynix all went down too.

When computer-parts stocks go down, the whole tech market often goes down. Investors are worried about how to feed the new AI computers with enough memory.

company

a group of people who work together to make or sell something

boss

the person who runs a company

stock

a small piece of a company that people can buy or sell

price

how much money you must pay for something

down

going to a lower level or number

computer

an electronic machine that does many tasks with information

memory

the part of a computer that holds information

AI

short for artificial intelligence — computer programs that can learn and answer questions

Level 2 — Elementary

Memory-chip and hard-disk-drive stocks fell sharply in the United States on Tuesday, May 19, 2026. The drop began after Seagate's chief executive, Dave Mosley, spoke at a Morgan Stanley investor conference and warned that AI data centres need so much more storage now that new factories cannot be built fast enough.

Mosley said that the company is fully sold out for the next several quarters. He added that 'any new fab takes too long' to deliver — meaning the time to design, build and start producing chips in a new factory is several years.

Investors reacted quickly. Seagate shares closed 6.8% lower. Western Digital, which also makes hard drives, fell 5.9%. Memory-chip makers Micron Technology dropped 6.2% in New York and South Korea's SK hynix closed 4.4% lower in Seoul.

Because chip makers are a very big part of the technology sector, the whole market felt the pressure. The Philadelphia Semiconductor Index, often called the SOX, finished the day 2.7% lower, and the broad S&P 500 index closed 0.5% lower.

memory chip

a tiny electronic part that stores information for a computer

hard-disk drive

a device that stores large amounts of computer information on spinning disks

chief executive

the most senior leader of a company; the CEO

investor

a person who buys parts of companies hoping to make money over time

conference

a meeting where many people come together to listen and talk about a subject

fab

short for 'fabrication plant' — a factory that makes computer chips

shares

small pieces of ownership in a company

index

a number that shows how a group of stocks is doing as a whole

Level 3 — Intermediate

Memory-chip and high-capacity-storage stocks led a broad Nasdaq technology sell-off on Tuesday, May 19, 2026 after Seagate Technology chief executive Dave Mosley told the Morgan Stanley Annual TMT Conference in San Francisco that hyperscale AI data-centre storage demand had moved 'so fast that any new fab takes too long' — projecting a multi-year structural supply deficit in nearline HDD capacity and bit-density-leading DRAM and HBM modules. Mosley reiterated guidance that Seagate is committed-capacity sold-out through fiscal Q3 2027 on hard-disk-drive shipments and added that the company has begun deferring smaller hyperscale orders in favour of long-term take-or-pay agreements with the three largest U.S. hyperscalers.

The remarks reverberated immediately across the storage and memory complex. Seagate closed 6.8% lower at $98.40, Western Digital fell 5.9% to $84.20, Micron Technology dropped 6.2% to $146.50 and Solidigm parent SK hynix closed down 4.4% in Seoul ahead of the U.S. open, with Kioxia and Samsung Electronics likewise off between 3% and 5% in Tokyo and Seoul. The Philadelphia Semiconductor Index (SOX) closed 2.7% lower at 6,490, the largest single-day decline since the early April Trump-tariff scare; Nvidia, the most-weighted SOX component, fell 1.4% in sympathy and clipped 0.5% off the broader S&P 500.

Sell-side analysts split on whether Mosley's framing is genuinely new news or simply louder packaging of the supply tightness already telegraphed in the most recent earnings cycles. Bank of America's Wamsi Mohan raised his fiscal-2027 HDD ASP forecast by 11% and his Seagate price target to $135; Morgan Stanley's own Joseph Moore moved Micron and SK hynix to overweight on the read-through to enterprise-SSD pricing; Bernstein's Stacy Rasgon, however, cautioned that the share-price reaction overshoots given that long-term take-or-pay contracts smooth pricing rather than amplify it.

The broader macro context matters: hyperscale capex for AI infrastructure is forecast by Goldman Sachs to grow another 38% in 2026 to roughly 470 billion dollars across Microsoft, Alphabet, Amazon, Meta and Oracle, with storage now the binding constraint in many incremental data-centre designs after power and cooling. Mosley's comments thus crystallise a thesis already familiar to specialist investors but newly priced by generalist money on a single trading day.

nearline HDD

the class of hard-disk drives that sit between online (always-spinning) and offline (tape-archived) storage, providing dense, less-frequently-accessed capacity at data-centre cost points

DRAM

dynamic random-access memory; the volatile main-memory technology used as system RAM and in graphics cards

HBM

high-bandwidth memory; stacked DRAM packages connected to a GPU or AI accelerator via silicon interposer to deliver very high memory bandwidth

take-or-pay agreement

a long-term supply contract under which the buyer pays for a committed minimum quantity whether or not it physically takes delivery

Level 4 — Advanced

Memory-chip and high-capacity-storage equities led a synchronised sell-off in U.S.-listed technology stocks on Tuesday, May 19, 2026 after Seagate Technology Holdings chief executive Dave Mosley told the Morgan Stanley Annual Technology, Media and Telecom Conference at the Palace Hotel in San Francisco that hyperscale AI data-centre storage demand had moved 'so fast that any new fab takes too long' — explicitly characterising the resulting nearline-HDD shortfall as a 'multi-year structural deficit' rather than a transient cyclical mismatch. Mosley reiterated guidance that Seagate is committed-capacity sold-out through fiscal Q3 2027 on Mozaic HAMR-bearing nearline HDD shipments and disclosed that the company has begun deferring smaller hyperscale orders in favour of long-term take-or-pay agreements with the three largest U.S. hyperscalers under terms running through calendar 2031.

The remarks reverberated immediately across the storage and memory complex. Seagate closed 6.8% lower at $98.40 on volume of 18.4 million shares (versus a 90-day average of 7.1 million); Western Digital fell 5.9% to $84.20; Micron Technology dropped 6.2% to $146.50; and SK hynix — the parent of NAND-flash supplier Solidigm and a critical HBM3E supplier to NVIDIA's Blackwell-Ultra family — closed 4.4% lower in Seoul ahead of the New York open. Kioxia Holdings and Samsung Electronics likewise traded off 3.2% and 4.6% respectively in Tokyo and Seoul. The Philadelphia Semiconductor Index (SOX) closed 2.7% lower at 6,490.18, the largest single-session decline since the early-April Trump-tariff scare; Nvidia, the most-weighted SOX component, fell 1.4% in sympathy and dragged the S&P 500 down 0.51% to 7,365.40 on broad volume.

Sell-side response was bifurcated. Bank of America Securities' Wamsi Mohan upgraded Seagate price target to $135 from $118 and raised his fiscal-2027 industry HDD ASP forecast by 11%, citing the Mozaic HAMR yield ramp's contribution to volumetric capacity rather than unit volume; Morgan Stanley's Joseph Moore lifted Micron and SK hynix to Overweight on the read-through to enterprise-SSD ASP momentum and DDR5/HBM3E pricing power; Bernstein Research's Stacy Rasgon, by contrast, published a same-day note cautioning that long-term take-or-pay structures smooth realised pricing rather than amplify it, that the cited 'multi-year fab build' is well-trodden ground for specialist investors, and that the share-price reaction overshoots fundamentals by roughly 250 basis points.

The broader macro context is the binding factor. Goldman Sachs Research's revised hyperscale-capex tracker, updated overnight on May 19, now projects calendar-2026 aggregate capital expenditure across Microsoft, Alphabet, Amazon, Meta and Oracle of approximately 470 billion dollars (up 38% year-on-year), with the marginal AI-training cluster increasingly bottlenecked on storage capacity rather than on accelerators, power-purchase-agreements or rack-level liquid cooling. Mosley's framing thus crystallises a thesis already in the price for specialist investors but newly internalised by generalist money on a single trading session — historically a setup that has preceded prolonged secular re-ratings rather than mean-reversion.