Micron Earned $24.67 Per Share Last Quarter. Its Dividend Is Still 15 Cents. Something Has to Give.

Micron posted record fiscal Q3 results with GAAP EPS of $24.67, up from $41.46B revenue (+346% YoY) and $28.24B net income, driven by AI high-bandwidth memory demand. Management guided to about $50B in revenue for the current quarter and said memory is booked out well into next year, supported by $25.4B operating cash flow and $18.3B adjusted free cash flow. While the dividend stays at just $0.15/share (payout ratio <1%), Micron plans to return 100% of excess cash to shareholders over time, likely shifting to larger capital returns later this year after capacity buildout (capex $7.1B in the quarter).

Analysis

The market should treat this less as a dividend story and more as a signal that AI memory has entered an allocation contest: cash is being pulled toward capacity before it is returned to owners. That favors the upstream semiconductor ecosystem in the near term, especially NVDA, because tighter HBM supply reduces the risk of accelerator shipment bottlenecks; the second-order effect is that AI compute vendors get more visible unit growth even if component economics stay concentrated at the memory layer. The losers are anyone assuming current excess cash is durable enough to support aggressive capital returns before the cycle matures.

The bigger risk is timing. In the next 1-3 months, the stock can keep rerating on buyback/dividend optionality and on any evidence that HBM demand remains booked out. Over 6-18 months, the key question is whether the current capex wave creates the usual memory oversupply lag; if it does, earnings power can fall much faster than the market expects, and the multiple will compress before the cash returns ever scale. What would falsify the bullish setup is a guide that implies capex stays elevated while pricing or gross margin momentum stalls.

Contrarian view: consensus is likely underweighting how little a token payout matters at the top of a cycle relative to discipline around wafer starts and capacity timing. The right trade is not to chase the yield narrative; it is to own the durability of AI memory demand while hedging the inevitable normalization risk. If management surprises with an earlier-than-expected repurchase authorization, the upside can extend; if not, this starts to look like peak-cycle cash conversion rather than a new capital-return regime.

Key Decisions for Investors

• Long MU on any 3-5% post-earnings pullback; 1-3 month upside is tied to a larger capital-return authorization or reaffirmation of HBM demand. Falsify if management signals capex remains structurally elevated into next year without margin protection.
• Pair trade: long MU / short a basket of high-multiple AI beneficiaries that depend on the same supply chain, using NVDA as the hedge leg only if market action starts treating the print as 'AI beta' rather than a memory-specific event. This captures relative upside if memory scarcity persists but AI multiples compress.
• Set a risk alert for any sign of DRAM/HBM pricing rollover or a gross margin inflection over the next 2 quarters; that is the earliest warning that current cash generation is peak-cycle and the stock should de-rate before reported earnings do.
• If MU gaps higher on buyback headlines, consider a call spread instead of outright stock to express the 1-3 month catalyst while capping exposure to the 6-18 month oversupply risk.