Evid Invest
EvidInvest
← Back to Blog

Micron's AI memory cycle is now visible in the filings

·EvidInvest Team
MUMicronAetherSEC filingsHBMAI infrastructuresemiconductors

Micron's AI memory cycle is now visible in the filings

This is financial research content, not investment advice. Built from Aether search over Micron SEC filings and checked against MU's FY2026 Q2 10-Q, FY2026 Q2 8-K earnings exhibit, and FY2025 10-K.

Run your own SEC filing analysis with Aether: aether.evidinvest.com maps filings into evidence labels so you can separate confirmed facts from inferred relationships.

Thesis

Micron's latest filings show that the AI memory cycle has moved from narrative into reported financials. Revenue, gross margin, earnings, and operating cash flow all reset sharply upward in FY2026 Q2, and management's FY2026 Q3 guide points to another step higher.

But the more useful conclusion is not simply "MU is an AI stock." Aether's SEC-first evidence stack shows three connected signals:

  1. Financial growth is now visible in the income statement.
  2. Customer demand is concentrated in cloud memory, but the key customer remains unnamed in the filings.
  3. Supplier-chain and capacity execution are central constraints for the HBM/AI-memory ramp.

That combination is exactly where SEC-first research matters. It separates confirmed filing facts from inferred relationships and keeps us from overclaiming named customers or suppliers the filings do not identify.

What changed

Micron reported FY2026 Q2 revenue of $23.86 billion, compared with $13.64 billion in the prior quarter and $8.05 billion in the same quarter last year. GAAP net income was $13.79 billion, or $12.07 per diluted share. Operating cash flow was $11.90 billion, compared with $8.41 billion in the prior quarter and $3.94 billion a year earlier.

The margin reset is just as important. GAAP gross margin reached 74.4% in FY2026 Q2, up from 56.0% in FY2026 Q1 and 36.8% in FY2025 Q2. Micron's 10-Q consolidated table shows the same operating leverage: revenue of $23.860 billion, gross margin of $17.755 billion, operating income of $16.135 billion, and net income of $13.785 billion.

Management's Q3 guide keeps the pressure on the thesis. Micron guided FY2026 Q3 revenue to $33.5 billion ± $750 million, with gross margin around 81% and non-GAAP EPS of $19.15 ± $0.40.

Evidence label: confirmed SEC fact for the reported Q2 results and management guidance. The caveat is that guidance is not realized revenue; it should be checked against the next filing before being treated as a sustained multi-quarter trend.

Financial growth: broad acceleration, data-center gravity

Micron's latest 10-Q shows strong growth across all four reported business units:

  • Cloud Memory Business Unit (CMBU): $7.749B in FY2026 Q2 revenue, up 47% sequentially.
  • Core Data Center Business Unit (CDBU): $5.687B, up 139% sequentially.
  • Mobile and Client Business Unit (MCBU): $7.711B, up 81% sequentially.
  • Automotive and Embedded Business Unit (AEBU): $2.708B, up 57% sequentially.

Year over year, the increases were even more dramatic: CMBU up 163%, CDBU up 211%, MCBU up 245%, and AEBU up 162% for the quarter versus FY2025.

The 2025 10-K explains the strategic driver. Micron says CMBU revenue increased 257% in FY2025 versus FY2024, primarily from DRAM bit shipments and higher average selling prices driven by accelerating AI demand in cloud server markets for HBM, high-capacity DIMMs, and low-power server DRAM. Micron also says it shifted DRAM supply toward data center and hyperscale cloud markets, with emphasis on HBM products.

Evidence label: confirmed SEC fact that AI, cloud server demand, HBM, high-capacity DIMMs, and low-power server DRAM are named growth drivers. The broader portfolio-pull effect is an inferred relationship: AI/HBM scarcity likely improves pricing, mix, and allocation across more than one segment, but that mechanism should not be stated as a precise quantified driver unless Micron quantifies it.

Customer and end-market signals: concentrated, important, unnamed

The customer signal is one of the most useful parts of the SEC evidence.

Micron's FY2026 Q2 10-Q says revenue from one customer was 13% of total revenue for the first six months of FY2026, primarily included in CMBU. In FY2025, one customer represented 17% of total revenue, also primarily included in CMBU.

That is a strong signal, but it has a boundary: Micron does not name the customer in these filings. We should not turn that into a named NVIDIA, hyperscaler, or OEM claim without a separate source.

What we can say from the filings is narrower and stronger: a double-digit customer concentration sits primarily in the cloud memory segment, the same segment tied to HBM and AI cloud-server demand.

The geographic signal matters too. Micron says roughly one-third of FY2025 revenue came from customers headquartered outside the United States, while roughly 80% of FY2025 revenue was from products shipped to customer locations outside the United States. In other words, the AI-memory thesis is global in both customer base and physical supply-chain footprint.

Evidence label: confirmed SEC fact for customer concentration and geographic exposure. The connection between the unnamed customer and AI/HBM demand is an inferred relationship; naming the customer remains needs source.

Supplier-chain and dependency signals: the upside has bottlenecks

Aether's supplier-chain lens is where the thesis becomes more operational.

Micron's 10-K says it generally has multiple sources of supply, but only a limited number of suppliers can deliver certain materials, components, and services that meet its standards. In some cases, supply can be single-source or sole-source, and new suppliers may not be qualified quickly.

The categories are specific: chemicals, silicon wafers, gases, photoresists, semiconductors, substrates, lead frames, printed circuit boards, targets, reticle glass blanks, third-party controllers, analog integrated circuits, outsourced foundry services, assembly/test providers, contract manufacturers, logistics carriers, advanced-equipment maintenance providers, and utilities.

This is not a list of named suppliers, and we should not pretend it is. But it is enough to show why the AI-memory ramp is not just about demand. HBM and leading-edge DRAM/NAND require advanced process technology, packaging capacity, equipment uptime, materials reliability, and qualification discipline.

Micron's capacity disclosures reinforce that point. The company describes investments in its global manufacturing footprint, including Taiwan modernization for DRAM/HBM products and Singapore HBM advanced packaging capacity expected to expand beginning in calendar 2027. In FY2026 Q2, Micron also disclosed the acquisition of a wafer fabrication facility in Tongluo, Taiwan, with expected meaningful shipments from the existing fab beginning in 2028 and plans for a second cleanroom by end-2026.

Evidence label: confirmed SEC fact for limited/single-source supplier categories and disclosed capacity moves. The view that HBM ramp complexity amplifies those constraints is an inferred relationship. Named critical suppliers remain needs source.

Risks and caveats

The filings support a strong operating thesis, but they also define the limits:

  • Customer concentration risk: one unnamed customer represented 13% of first-half FY2026 revenue and 17% of FY2025 revenue, primarily in CMBU.
  • Supplier qualification risk: some inputs and services can be limited-source, single-source, or sole-source.
  • Capacity timing risk: Taiwan fab modernization, HBM packaging expansion, and future shipments are multi-year execution items.
  • Geographic exposure: approximately 80% of FY2025 revenue was from products shipped to customer locations outside the United States.
  • Evidence boundary: SEC filings do not name the major customer or critical suppliers in the retrieved evidence.
  • Forecast boundary: FY2026 Q3 guidance should not be treated as realized results.

This is why the right framing is not a buy/sell call. It is an evidence map: what Micron has confirmed, what Aether can connect, and what still needs source verification.

How Aether changes the workflow

A normal search for MU and AI produces a market narrative. Aether's SEC-first workflow produces a cleaner evidence stack:

  • 10-Q: current financials, segment revenue, customer concentration, geographic exposure, and updated risk language.
  • 8-K / Exhibit 99.1: earnings release, management quote, guidance, and operating metrics.
  • 10-K: annual business-unit context, HBM/AI demand language, customer concentration, global shipment exposure, and supplier-chain risk factors.
  • DEF 14A / SD / other 8-Ks: available for governance, conflict-minerals, or capital-structure context, but not central to this operating thesis.

That distinction matters for financial content. The best claims are not the loudest claims. They are the labeled claims:

  • Confirmed SEC fact: Q2 revenue, margin, EPS, cash flow, guidance, segment growth, one-customer concentration, global shipment exposure, limited/single-source supply categories, and disclosed capacity expansion.
  • Inferred relationship: AI/HBM demand is likely driving higher-value customer concentration and supply allocation.
  • Needs source: named customers, named critical suppliers, and whether Q3 guidance becomes realized performance.

Bottom line

Micron's filings show an AI memory cycle that has become financially visible. Revenue and gross margin reset upward, cloud/data-center memory is central, one unnamed CMBU-heavy customer accounts for double-digit revenue share, and supplier-chain execution remains a key constraint.

For investors and operators, the useful question is not simply "is MU an AI stock?" The better question is: which parts of the AI memory supply chain are confirmed in the filings, and which parts are still inferred?

That is the gap EvidInvest and Aether are built to close.

Do your own Aether analysis

Want to test the same workflow on another company? Use aether.evidinvest.com to search SEC filings, map financial/customer/supplier-chain evidence, and label claims as confirmed, inferred, or needing a source.

Source notes

Fair Value Weekly

Get DCF breakdowns, fair value updates, and portfolio ideas for serious investors. No spam, no paywalled teasers.