Conviction / Shovels
MU
Micron Technology
conviction
Micron makes the specialised memory that advanced chips cannot run without, and it has sold out its 2026 supply at agreed prices while posting record revenue and margins near 85%. The open question is not demand but duration, because margins that high are usually a peak, and the share price is discounting exactly that.
What we judge
The view on expected return from here, 0 to 100.
Whether the thesis could succeed without this company at all. This moves slowly, and only when the company's position genuinely changes.
What we measure
Facts from market data, not judgments.
| Price | $923.40 |
| 30-day move | -1.6% |
| Position in 52-week range | 71 |
| From the 52-week high | -26.4% |
| Average daily volume | $45.7B |
Why it is in the fund
Thesis
HBM and high-capacity DRAM are critical bottlenecks for agentic AI workloads that demand massive memory bandwidth.
Standing judgment
About 27% off the high at roughly 6x forward earnings. That multiple is not cheapness, it is the market declining to capitalise peak memory pricing, with CXMT adding DRAM supply. This is structurally critical and cyclically unresolved at the same time, which is exactly the 85-89 description.
What changed, and why
Every move in this number carries the reason it moved. A score does not change without new evidence.
| Date | Conviction | Change | Cause | Evidence |
|---|---|---|---|---|
| 2026-08-05 | 87 | +9 | High bandwidth memory is a hard physical bottleneck and the reported numbers are the strongest in the fund, but the memory cycle is the live risk that keeps this out of the top band. | FQ3 26 reported 2026-06-24: revenue $41.5B (+346% y/y, +74% q/q), gross margin 84.9%, non-GAAP EPS $25.11, data center about $11.5B versus $1.53B a year earlier. FQ4 guided to roughly $50B revenue at about 86% gross margin. HBM4 has been shipping for Vera Rubin since March at roughly twice the ramp |
| 2026-08-05 | 78 | flat | first v3 judgment (v2 score was uninformative, see L-001): Record earnings power is real and sold out through 2026, but memory margins at this level are the signature of a cycle top and the multiple will compress against them, not expand. | FQ3 2026 revenue $41.46B, up 346% year over year and 74% sequentially, fifth consecutive record; non-GAAP gross margin 84.9%, a company record, versus 39.0% a year ago; non-GAAP EPS $25.11 versus $20.71 consensus; FQ4 guide $50.0B plus or minus $1.0B at roughly 86% gross margin; entire 2026 HBM3E an |
| 2026-08-05 | 78 | +57 | first v3 judgment (v2 score was uninformative, see L-001): Record earnings power is real and sold out through 2026, but memory margins at this level are the signature of a cycle top and the multiple will compress against them, not expand. | FQ3 2026 revenue $41.46B, up 346% year over year and 74% sequentially, fifth consecutive record; non-GAAP gross margin 84.9%, a company record, versus 39.0% a year ago; non-GAAP EPS $25.11 versus $20.71 consensus; FQ4 guide $50.0B plus or minus $1.0B at roughly 86% gross margin; entire 2026 HBM3E an |
| 2026-07-26 | 21 | -14 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-25 | 35 | +17 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-24 | 18 | -6 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-23 | 24 | +3 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-22 | 21 | -17 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-21 | 38 | +14 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-20 | 24 | +17 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-19 | 7 | -2 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-18 | 9 | +8 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-17 | 1 | -6 | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-16 | 7 | flat | migrated from v2 (cause not recorded) | v2 `scores` row |
| 2026-07-15 | 7 | -11 | migrated from v2 (cause not recorded) | v2 `scores` row |