Tech stocks surge as Micron earnings ease AI fears, oil falls further
Micron's fiscal results showed revenue of $37.4B (+48.9% YoY) with $7.59 diluted EPS and 39.8% gross margins, easing fears that AI-driven memory demand was softening. The strong print reframes the AI infrastructure buildout narrative as intact, creating a setup where memory-cycle bulls re-engage while bears point to peak-margin risk.
What happened
Micron's fiscal results showed revenue of $37.4B (+48.9% YoY) with $7.59 diluted EPS and 39.8% gross margins, easing fears that AI-driven memory demand was softening.
The read · Jun 25
MU's near-50% revenue surge and expanding margins raise the question of whether the AI memory cycle has more room to run or whether these numbers represent peak-cycle conditions that won't repeat in FY2026.
Wrong if
Memory cycles historically mean-revert sharply; if hyperscaler inventory builds or DRAM spot prices soften in the next 4-6 weeks, consensus estimates get cut and the multiple compresses quickly regardless of the strong FY2025 base.
CoverageSource: Investing.com · Published here THU, JUN 25 · 9:16 AM ET · 2 outlets in this record · latest listed: Yahoo Finance at 9:16 AM ETHow this is decided →
The chart · MUTradingView · third-party feed, not the Wire’s licensed closes
🔒 Click to interact · scroll moves the page
The story1 min read
Micron Technology reported fiscal 2025 results with revenue of $37.4B, up nearly 49% year-over-year, alongside diluted EPS of $7.59 and gross margins of 39.8%. The print landed well enough to broadly lift tech sentiment, with investors reading the numbers as confirmation that hyperscaler AI spending is still translating into real memory demand rather than slowing.
The results matter because Micron is widely viewed as a real-time demand proxy for the entire AI infrastructure cycle — HBM and DRAM volumes tied to GPU clusters make MU one of the most direct reads on whether AI capex is genuinely absorbing supply. A near-50% revenue jump with improving net margins (22.8%) is not a soft beat.
The bull tension here is whether this is cycle peak or cycle mid-point. At these margin levels, bears will argue that DRAM pricing is mean-reverting and that the memory industry's historical boom-bust pattern makes 39.8% gross margins unsustainable into FY2026. The counter is that HBM supply remains structurally tight and that MU is still ramping its HBM3E capacity, which commands pricing well above commodity DRAM.
What to watch: forward guidance language on HBM pricing, any commentary on customer inventory levels at major cloud hyperscalers, and whether the broad tech lift holds into the next session or fades as a one-day sentiment bounce.
The read, in fullWhy it matters
MU posted $37.4B in revenue (+48.9% YoY) with 39.8% gross margins and $7.59 diluted EPS — numbers that directly refute near-term AI demand fears. HBM3E ramp is still in early innings, and if hyperscaler capex holds, pricing power in high-bandwidth memory keeps margins elevated through at least the first half of FY2026. The post-earnings tech sector lift suggests sentiment was underpositioned into the print.
The arc● Open · no developments yet
· Published
The read above, as written. kept as written · closes shown from JUN 25 on
Next · what would move this
4-6 weeks. Follow to be told when one lands.
Since this story · named here, equal weight · 1D EOD-5.0%
JUN 25 · first close after publicationSEP 4
Price context does not establish that the story caused the move.
▲ The case it holds
With HBM3E supply still constrained and FY2025 revenue growing nearly 49% to $37.4B at 39.8% gross margins, MU's results suggest the AI-driven memory upcycle has pricing discipline that prior DRAM cycles lacked, leaving room for further estimate revisions higher.
▼ The case it breaks
Memory semiconductors have a well-documented boom-bust history, and 39.8% gross margins near multi-year highs historically signal a cycle top rather than a new floor — any softening in cloud capex or customer inventory digestion could compress margins and EPS sharply into FY2026.
Kept as written · your side, if you take one, is graded privately against licensed closes after 10 trading days · nothing here is advice · How the Wire is made →
Shares a name with this story — discovery, not a connection.
The memory
This page is kept as it was written on Jun 25. Later coverage joins it only when the company and catalyst evidence match, and what the stock did is shown from licensed end-of-day closes — never re-graded, never backdated. The judgment is yours.