Semiconductor Intelligence analyst Bill Jewell’s trade data analysis<\/a>. The mechanism is direct: Samsung, SK Hynix, and Micron, which control 95% of global DRAM production, have systematically converted fab capacity toward high-bandwidth memory for AI accelerators, which earns three to five times more revenue per wafer than consumer DRAM. PC shipments fell 4.9% and smartphones dropped 6.7% in Q2 2026, not from weak demand but from constrained supply. Dell, HP, and Apple all said so on earnings calls.<\/p>\nWhat this means for your business<\/h2>\n
The enterprises least exposed to this shortage are the ones that either locked in memory procurement contracts 12 to 24 months ago or whose device architectures keep memory as a small share of total bill of materials. Apple gained PC market share to nearly 10% precisely because it did both. Companies that left device refresh cycles on auto-pilot, relying on spot procurement or short-term supplier relationships, are the ones now watching HP absorb a 9% shipment decline while Lenovo’s scale insulates it down to 2.1%. Your position in this shortage is largely a function of decisions made in 2024.<\/p>\n
The structural argument here is worth taking seriously, even though Jewell publishes through SemiWiki, a platform with close ties to semiconductor industry clients who benefit from narratives that emphasize supply-side constraints over demand softness. The data he cites are drawn from US Customs records and IDC trackers, not proprietary estimates, and the earnings disclosures from Dell, HP, and Apple are on the record. The 3:1 conversion ratio, meaning each high-bandwidth memory wafer displaces three conventional DRAM wafers, comes from Micron’s own public disclosures. The frame doesn’t corrupt the numbers here.<\/p>\n
What the CTO community tends to underweight is that new fab capacity, when it arrives in 2027 and 2028, doesn’t automatically flow back to consumer or enterprise device markets. Samsung, SK Hynix, and Micron will face the identical revenue calculation that drove the reallocation: high-bandwidth memory for hyperscalers pays more per wafer than LPDDR5X for phone OEMs. IDC calls the current reallocation “potentially permanent.” The falsification condition for that claim is a significant cooling of hyperscaler AI capex, which would shift the per-wafer math, but every major cloud provider’s public guidance points the opposite direction through at least 2027.<\/p>\n
For CTOs managing device refresh budgets, the certified refurbished market for 2024 and early 2025 hardware carries pre-shortage memory cost structures and represents genuine value that new 2026 devices won’t match at equivalent price points. The more consequential question isn’t what to buy this quarter. It’s whether your procurement function has the supplier relationship depth to secure memory-intensive components at contracted prices before the next capacity wave creates the next allocation crunch, because this is now the recurring failure mode in enterprise hardware planning.<\/p>\n
Concept deep-dive: High-bandwidth memory<\/h2>\n
High-bandwidth memory, or HBM, is DRAM built in vertical stacks, up to 16 chips tall, bonded directly onto a processor die rather than placed separately on a circuit board. Think of conventional DRAM as books on a shelf you have to walk to; HBM is the same books sitting in your lap. The stacking is physically expensive to manufacture, consuming three to four times the fab capacity per usable memory bit compared to standard DRAM, which is exactly why converting to HBM production crowds out the chips that go into phones and laptops.<\/p>\n
Based on reporting from AI Chip Imports Eclipse PC and Phone Shipments as Memory Shortage Squeezes Consumers<\/a>, originally published 2026-07-24 11:10:00.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"Share with your CTO AI chip imports into the US hit $165 billion in 2025, surpassing the combined $109 billion in smartphone and PC imports, according to Semiconductor Intelligence analyst Bill Jewell’s trade data analysis. The mechanism is direct: Samsung, SK Hynix, and Micron, which control 95% of global DRAM production, have systematically converted fab […]<\/p>\n","protected":false},"author":1,"featured_media":6498,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[147],"tags":[207],"tmauthors":[],"class_list":["post-6497","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai-infrastructure","tag-cto"],"_links":{"self":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts\/6497","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/comments?post=6497"}],"version-history":[{"count":0,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts\/6497\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/media\/6498"}],"wp:attachment":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/media?parent=6497"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/categories?post=6497"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/tags?post=6497"},{"taxonomy":"tmauthors","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/tmauthors?post=6497"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}