Micron says AI-driven reminiscence demand continues to outstrip provide by 2026, with lean DRAM inventories and prospects solely assembly a fraction of their wants, holding business circumstances tight.
The corporate is increasing capability through node transitions and greenfield initiatives — notably the 1-gamma DRAM ramp, Idaho One (mid‑2027), the Tongluo web site (closing anticipated Q2, provide late‑2027/2028) and a brand new Singapore NAND fab (first wafer H2 2028) — that means provide constraints might prolong into 2028.
Micron experiences HBM4 is in excessive‑quantity manufacturing with buyer shipments underway, yields on observe, and calendar Q1 shipments ramping sooner than beforehand guided, whereas 2026 HBM capability is already offered out.
Excited by Micron Expertise, Inc.? Listed here are 5 shares we like higher.
Micron Expertise (NASDAQ:MU) executives advised traders at Wolfe’s Auto, Auto Tech, and Semiconductor Convention that enterprise circumstances have strengthened because the firm’s most up-to-date earnings name, with demand persevering with to outstrip provide throughout reminiscence markets and tight business circumstances anticipated to persist past 2026.
Chief Monetary Officer Mark Murphy, joined by Head of Investor Relations Satya Kumar, stated Micron is seeing “considerably increased” demand than it—or the business—can provide. Murphy added that the corporate is working to increase capability by a mixture of node transitions and greenfield investments, whereas additionally progressing on multi-year provide agreements with prospects searching for longer-term visibility and assurance.
→ Is Albemarle Setting Up for a Lithium-Fueled Rebound?
Murphy stated demand is being pushed by AI programs that require “extra and higher reminiscence” as fashions develop bigger, context home windows prolong, and reasoning turns into extra intense. He described rising “tokens generated” and a broader “re-architecture of reminiscence programs” that’s pushing high-performance reminiscence deeper into system architectures.
He pointed to bettering server demand tendencies by 2025, noting that expectations moved from single-digit development to the “mid-teens” as AI workloads expanded and commenced affecting even conventional server demand. Murphy additionally cited rising hyperscaler capital spending, saying expectations for 2026 are “near $800 billion,” in contrast with “beneath $200 billion” just a few years in the past.
→ Cloudflare: One other AI-Disrupted Inventory You May Need to Purchase
Whereas information heart stays the first driver, Murphy stated demand is predicted to proliferate to the sting over time, together with smartphones and PCs, and finally areas tied to autonomy-related actions.
On provide, Murphy stated Micron’s inventories are lean—significantly in DRAM—and the corporate is “sweating the property” to supply incremental bits the place doable. Even so, he emphasised that offer is just not assembly demand “by a considerable margin.” He referenced feedback from Micron’s CEO on a previous earnings name that some key prospects have been solely capable of meet 50% to two-thirds of their demand, underscoring how widespread the shortfall is.
→ AI Energy Crunch: Why Bloom Vitality Is the Hidden Winner
Murphy outlined a number of steps underway to increase capability over time:
Node transition: The corporate is ramping its 1-gamma DRAM node, which Murphy stated will present provide into calendar 2026 and is predicted to characterize the vast majority of Micron’s bits within the second half of calendar 2026.
Idaho One: Incremental greenfield capability is predicted to come back on-line in “mid-2027.”
Tongluo, Taiwan acquisition: Micron introduced an acquisition of a web site in Tongluo to help DRAM manufacturing. Murphy stated the transaction is predicted to shut in calendar Q2 and that the fab might present provide close to the top of 2027 into 2028. He additionally stated the acquisition value disclosed within the press launch was $1.8 billion and that the deal, together with software set up and associated spending, will add to capital expenditures.
Singapore NAND fab: Micron broke floor on a brand new NAND fab in Singapore, with “first wafer out” anticipated within the second half of 2028.
Murphy stated provide additions take time, significantly for greenfield capability. He additionally famous that node transitions are “much less environment friendly than it was,” limiting how a lot incremental provide the business can generate with out constructing new amenities.
One other issue pressuring provide, Murphy stated, is the rising silicon depth of high-bandwidth reminiscence (HBM). He described a commerce ratio that has traditionally been round 3-to-1 for HBM3 and stated it will increase because the business strikes to HBM4, HBM4E, and HBM5, additional tightening provide.
Murphy additionally addressed what he described as “current inaccurate reporting” about Micron’s HBM4 place. He stated Micron is in high-volume manufacturing of HBM4, has commenced buyer shipments, and expects cargo volumes to ramp efficiently in calendar Q1—one quarter sooner than the corporate had talked about on its December earnings name.
Murphy stated Micron’s HBM capability ramp is progressing and reiterated that the corporate has offered out its calendar 12 months 2026 HBM provide. He added that HBM yields, together with HBM4 yields, are on observe, and stated the HBM4 product delivers over 11 Gbps speeds. He expressed confidence in efficiency, high quality, and reliability.
Requested about HBM share, Murphy stated the corporate beforehand focused ramping share in step with its standard DRAM share and said that Micron achieved that. He added that HBM is now managed like different merchandise with portfolio choices primarily based on buyer wants, manufacturing footprint, and worth.
Murphy reiterated that Micron had guided to 68% gross margin and stated the corporate’s monetary outlook has improved since earnings, pushed by pricing. He cautioned that at increased margin ranges, incremental value will increase have a smaller mathematical impression on margin enlargement than they do at decrease margins, however stated Micron expects margins to increase from the fiscal second quarter into the third quarter.
Along with pricing, Murphy cited robust value efficiency from the 1-gamma ramp, operational execution, “superb” spend management, and improved absorption from increased volumes. He additionally pointed to combine as a contributor, saying Micron’s premium expertise and portfolio present flexibility to optimize towards higher-value areas over time.
On NAND, Murphy stated the enterprise stays essential and described how the business had “ample provide” from late 2022 onward, prompting Micron to cut back utilization and gradual node migrations to revive supply-demand stability. Over that interval, he stated Micron sharpened its NAND portfolio, introduced controller functionality in-house, and prolonged course of expertise management.
Murphy stated Micron now holds a “preeminent place” in NAND, centered on the information heart SSD market, spanning efficiency TLC SSDs and capacity-oriented QLC drives. He stated Micron has gained share and famous that the corporate mentioned attaining “over a billion-dollar run fee once more” in NAND through the prior quarter.
Murphy described tighter NAND market circumstances as storage turns into a extra essential a part of AI system structure. He outlined a “scorching, heat, chilly” tiering strategy, with HBM serving the most well liked information, LPDRAM supporting heat information, and drives serving cooler information, together with rising key-value (KV) cache offload wants. He stated this dynamic contributed to Micron’s determination so as to add greenfield NAND capability in Singapore, with output anticipated in 2028, whereas provide wants earlier than then will probably be met by node transitions, together with a G9 transition.
Kumar added that Micron printed a white paper on LPDDR, describing programs anticipated to ship this 12 months with triple the LPDDR content material in contrast with final 12 months, and asserting that further LPDDR content material can cut back time to first token by 98% and considerably enhance inference.
Murphy closed by emphasizing that Micron will proceed investing in a “disciplined manner,” reassessing market circumstances and aggressive capability additions, whereas working to increase provide and safe longer-term agreements with prospects searching for provide assurance over multi-year horizons.
Micron Expertise, Inc is a worldwide semiconductor firm that designs and manufactures reminiscence and storage options. Its product portfolio contains dynamic random-access reminiscence (DRAM), NAND flash reminiscence, solid-state drives (SSDs), reminiscence modules and embedded reminiscence options for a variety of computing and digital gadgets. Micron provides elements utilized in information facilities, enterprise and cloud infrastructure, shopper computing, cellular gadgets, automotive programs and industrial functions, and in addition markets consumer-facing merchandise beneath the Essential model.
Based in 1978 and headquartered in Boise, Idaho, Micron has grown into a global producer with analysis, growth and manufacturing amenities throughout a number of areas.
The article “Micron Expertise Says AI Reminiscence Demand Nonetheless Outstrips Provide By 2026, HBM4 Transport Early” was initially printed by MarketBeat.
