MU Trans: Supply-demand tight through 2028; bigger capital returns from Dec.

Compiled by Dolphin Research:$ Micron Tech.US FY26 Q4 Earnings Call Trans

Earnings take:《Micron: Without a mania-driven spike, can LT agreements really carry the memory upcycle?》

I. Core takeaways

1. Capital return

a. From Dec 9, 2026 (two years after the final CHIPS agreement), capital return will step up, with a LT goal to return 100% of excess cash to shareholders, primarily via buybacks. b. Cash is expected to reach the target level by end-FQ1; $2.2bn remains under the current buyback authorization, with an additional authorization to be requested shortly.

2. Guidance

a. FQ1: revenue of $61.5bn ± $1.5bn, GPM around 86.25%, OpEx approx. $2.06bn, EPS of $38.15 ± $1 (based on ~1.13bn shares), and tax rate around 15.5%. b. FY27: sequential revenue growth each quarter, with a new full-year record. FQ1 is the GPM trough for the year, then margins trend higher, while price hikes moderate. Full-year OpEx to rise by approx. $2.5bn, driven mainly by R&D and higher incentive comp; tax rate around 15.5%. c. CapEx (net of Gov. incentives): ~$11.5bn in FQ1, approx. $25bn in 1H, higher in 2H (i.e., >$50bn, better than last quarter’s '>$10bn per quarter'). Fab construction CapEx will grow much faster than equipment.

3. GPM drivers: FQ4 GPM expanded 210bps QoQ, led by pricing and execution, partly offset by mix. FY26 incentive comp raised in FQ4, with most manufacturing-related portions capitalized into inventory and the higher-cost inventory largely sold in FQ1. In FQ2, this effect eases but is offset by higher FY27 incentive comp.

4. Key financials this quarter

a. Totals: FY26 revenue $133.2bn (+256%), GPM 81.1% (+40ppt), EPS $75.52; FQ4 OPM 82.3%, EPS $33.42. b. OpEx and cash flow: FQ4 OpEx $2.6bn (+$1.1bn QoQ), mainly due to across-the-board incentive comp increases and $300mn community investments; FCF $33.2bn. SCA customer prepayments are in financing cash flows and do not affect FCF, with $12.3bn received in the quarter. c. Balance sheet: DIO 129 days (+9 days, reflecting EOL-related stocking and incentive comp capitalized in inventory), expected to decline over coming quarters. Net cash $68.3bn, debt $5.2bn, and all three rating agencies raised ratings to BBB+. d. SCA basis: RPO approx. $150bn, counting only SCAs with a defined pricing framework and measured at committed volumes and floor prices, a conservative measure. Actual revenue over the contract terms should far exceed RPO, and even at floors, margins would be well above any prior cycle peak.

II. Detailed call notes

2.1 Key management commentary

1. Industry supply/demand

a. Demand has strengthened further since last call. CY27–CY28 memory supply/demand is expected to be meaningfully tighter vs. CY26, and even with new cleanrooms and upward revisions from customers, management sees no clear timing for equilibrium. b. DRAM: CY26 industry bit shipment growth around mid-20s%, Micron in line with the industry. CY27–CY28 to grow low-20s% with supply constrained both years; HBM bit growth to outpace conventional DRAM through CY28. c. NAND: CY26 industry bit shipment growth low-20s%, slightly above prior view, with Micron growing below industry. CY27–CY28 to grow mid-20s% with supply constrained both years. d. As a result, FY27 CapEx is raised, with most of the increase for fab construction to accelerate cleanroom availability post-2H CY28; existing cleanrooms will be optimized for higher output, and some equipment spend will be pulled forward.

2. Strategic Customer Agreements (SCA)

a. 26 multi-year take-or-pay deals signed (vs. 16 last quarter), expected to cover 35%+ of revenue through 2030. Three-quarters have a defined pricing framework (mostly with price bands), while the rest are periodically priced at market. b. New agreements extend to 2031, with two existing agreements extended by one year to 2031; newly negotiated pricing is being set higher, reflecting current market conditions. c. Customer financial commitments rose to $32bn, the vast majority in cash prepayments. Prepayments are unrestricted and will be gradually returned in the latter halves of the contracts, assuming minimum purchase volumes are met.

3. Data center

a. Server shipments are expected to grow high-teens% in both CY26 and CY27. Under memory tightness, per-box content growth will be slightly below prior expectations. b. HBM: FQ4 revenue grew faster than the company average. For CY27, the vast majority of bits are already contracted at significantly higher prices, narrowing the margin gap vs. conventional DRAM; Micron will co-develop with NVIDIA the industry’s first custom HBM4E (NVHBM) for next-gen GPUs and the NVLink Fusion platform. c. Data center SSD: FQ4 revenue was nearly $10bn, up more than 10x YoY, accounting for over two-thirds of NAND revenue. KV cache offload and HDD displacement are expanding the SSD market, and CY26 should mark a fifth straight year of share gains.

4. PC/handset and physical AI

a. Overall PC and handset unit volumes may decline double digits, but high-end demand should still drive industry revenue growth. Nearly half of FQ4 MCBU revenue came from 1-gamma products. b. L4+ autonomous driving typically requires >200GB memory and several TB of storage, both an order of magnitude above L2+/L3; humanoid robots are similar. Physical AI could become a major demand driver by decade-end, with multiple customers sampling next-gen products.

5. Technology and capacity

a. 1-gamma DRAM and G9 NAND are already the largest volume nodes. Next-gen DRAM and NAND nodes are planned for mass production in 2H CY27. b. DRAM capacity: Idaho ID1 to start wafers mid-CY27, ID2 by end-CY28; Japan expansion to start wafers by end-CY28; Tongluo, Taiwan to ship at scale by mid-CY27; New York’s first fab to start wafers in CY30. c. Singapore: HBM packaging to ship from early CY27 (cleanroom ahead of schedule), and the new NAND fab to start wafers in 2H CY28. After wafer start, it will take several quarters to reach effective output. d. People: Manish Bhatia promoted to President & COO, overseeing BUs, global ops, and P&L; Scott DeBoer promoted to President & Chief Technology & Product Officer.

2.2 Q&A

Q: What are the milestones for capital return? What is the minimum cash balance — keep $10bn and return the rest?

A: Cash will reach the target level by end-FQ1, after which excess cash will be returned primarily via buybacks, with no target amount disclosed. Under FQ1 guidance and planned CapEx, FQ1 FCF will be well above FQ4’s $33.0bn. Even with higher R&D and CapEx, FCF should grow strongly and more sustainably, supported by SCAs. The pace of return will depend on multiple factors, with higher returns starting Dec 9. The current buyback authorization has $2.2bn remaining, and the company will seek an increase soon.

Q: FY27 CapEx looks $55bn+, about high-teens% of revenue. Will 20%–25% capital intensity be the new normal LT?

A: No LT capital intensity target given; current intensity is at historical lows, and the industry has structurally reset. The spend mix is shifting toward fab construction, and that trend will continue for years; low capital intensity reflects memory as a strategic asset. The company will add capacity with discipline to ensure adequate returns on investments.

Q: How large is the high-cost inventory impact in FQ1 GPM guidance? Any other mix shifts?

A: About $1bn of extra costs in FQ1 from incentive comp and new fab ramps, with incentive comp the main driver. The FQ4 incentive comp uplift is most visible in OpEx, with most of the manufacturing portion capitalized into inventory, thus little FQ4 GPM impact and flowing through as higher-cost inventory sold in FQ1. FY27 incentive comp is also raised materially, first appearing in OpEx in FQ1, and the manufacturing portion impacting GPM starting in FQ2. Beyond normal cost increases from volume and D&A, there is about $1bn of ongoing costs in FY27 (incentive comp, start-up, etc.), some of which are variable or transitional; FQ1 is the GPM trough, with margins rising thereafter on continued pricing and execution.

Q: Do HBM price increases take effect on Jan 1? How does HBM growth compare?

A: Price increases begin in CY27, with most HBM supply already sold out at prices well above 2026 levels; no growth rate disclosed. CY26 prices reflected last year’s negotiations. At the industry level, HBM demand is growing faster than DRAM; the roadmap includes HBM3E and HBM4, with HBM4E by the end of next year.

Q: At the current pace, FQ2–FQ4 could generate over $100bn of cash. Why not return a similar amount in FY27?

A: No comment on that specific math; management reiterated both the capacity and intent to increase returns. The company will seek a higher buyback authorization and, under the CHIPS agreement, step up capital returns starting Dec 9.

Q: With new capacity and some customer down-speccing, can industry pricing hold or even exceed 2027 levels in 2028?

A: Management expects CY27–CY28 supply/demand to be tighter than CY26, supportive of pricing, with no specific price commentary. Supply: long build times for cleanrooms, gradual ramps after wafer start, and rising trade ratios moving to HBM4/4E, with diminishing per-wafer output gains at future nodes; even with some new cleanrooms in 2028, supply remains tight. Demand: some per-server content growth is slightly below expectations to enable more unit shipments; server units are seen up high-teens% in CY26 and CY27, laying the groundwork for 2028 DC DRAM demand. Over 75% of CY27 output is already locked, and most customer discussions are shifting to 2028. Both SCA and non-SCA customers are adding demand and want to extend coverage beyond 2030.

Q: A large customer is down-speccing HBM. If its capacity shifts back to DDR, will DDR supply jump?

A: No direct comment on that customer; reiterated that HBM demand growth outpaces conventional DRAM through CY28. Memory remains tight in CY27 and CY28. Customer optimization aims to ship more units and expand AI deployments, without diminishing the system’s latent need for more memory, and the marginal benefit of further optimization is declining; the overall demand trend remains healthy vs. LT memory needs.

Q: Is the heavier fab construction CapEx due to constrained tool availability next year?

A: No; it reflects severe supply/demand imbalance and long lead times for new capacity, not tool constraints. SCA-driven visibility supports new builds at ID1/ID2, Japan, Singapore, and Tongluo, with equipment deployed in line with updated demand.

Q: If all SCAs under negotiation get signed, can LT revenue coverage reach 60%–70%?

A: The goal remains about 50% of revenue covered by SCAs through 2030, potentially lower. The final ratio depends on the scale of other businesses. The prior comment that over 75% of CY27 output is locked includes both SCA and non-SCA customers, as some large customers operate on annual agreements and non-SCA customers have already placed 2027 orders. The company intentionally preserves flexibility to allocate supply across customers, end-markets, and new accounts. SCAs provide LT demand visibility, enabling better investment planning and dampening past industry volatility.

Q: Does KV cache offload to flash change NAND R&D and CapEx priorities?

A: Investment in NAND is increasing, with both R&D and manufacturing being ramped. DC SSD share has risen for five straight years, with FQ4 revenue at $10bn and two-thirds of NAND revenue. Singapore’s new NAND fab remains on track to start wafers by end-CY28, with CapEx also supporting G9 migration, equipment efficiency, and NAND R&D.

<End of text>

Risk disclosure and statement:Dolphin Research Disclaimer and General Disclosures