Here’s something worth sitting with for a moment: Micron Technology just reported the best quarter in its 48-year history — revenue up 4.5x year over year to $41.46 billion, non-GAAP EPS of $25.11 beating consensus, and a company-record non-GAAP gross margin of 84.9% — and the stock has since lost roughly a quarter of its value. The Q4 guidance called for $50.0 billion (± $1.0 billion) in revenue with gross margin approximately 86%. The stock peaked at $1,255 on June 25. As of this morning, it trades near ~$806.

That disconnect is the trade.

But it doesn’t exist in isolation. In 13 days, Amazon reports Q2 2026 earnings on July 30 with a roughly $200 billion capex expectation hanging over every analyst model on the Street. And the way those two events are connected tells you more about the current state of AI infrastructure investing than almost anything else happening in the market right now.

The Micron Paradox — By the Numbers

Start with what the tape is pricing versus what the business is doing. Micron’s fiscal Q3 (ended May 28, 2026) delivered $41.46 billion in revenue — up from $9.30 billion a year earlier and $23.86 billion the prior quarter. That’s a 74% sequential jump in a single quarter. Non-GAAP EPS hit $25.11, and the company guided Q4 for approximately $50.0 billion in revenue with a gross margin of approximately 86%.

The HBM supply situation is structurally constrained. Micron has said it signed 16 strategic customer agreements (SCAs) in fiscal Q3, structured as take-or-pay agreements with binding volume commitments. Fourteen of those 16 SCAs represent approximately $100 billion in cumulative minimum contracted revenue over the remaining agreement term. Micron also disclosed $22 billion in cash deposits and other financial commitments associated with these SCAs.

And yet the stock is down ~25% in less than three weeks. The 52-week range is $103.38 to $1,255. At ~$806, it trades at roughly 19x trailing earnings and under 7x forward earnings — one of the widest forward-versus-trailing discrepancies in the entire semiconductor sector.

So what broke the tape? Three things converged in roughly a week.

First: high-profile investor Michael Burry disclosed a put-options position in Micron (via a Scion Asset Management filing), which added visible negative sentiment. Second: China’s ChangXin Memory Technologies (CXMT) — the largest domestic DRAM manufacturer — set terms for a Shanghai IPO expected to raise roughly $8.5–$8.6 billion (about RMB57.9 billion), with proceeds tied to scaling efforts. Third, and most immediately damaging: reports emerged mid-July that the U.S. government is considering new export restrictions affecting advanced memory and related technology — which hit sentiment around HBM exposure.

None of these catalysts changed Micron’s actual operational performance for the current quarter. But they converged on the same nerve: the bull case for MU relies on a sustained HBM pricing supercycle, and two of the three shocks were targeted directly at that assumption.

What the Numbers Actually Say About Risk

Here’s where it gets interesting. The bear case is real but it’s also been priced in faster than the fundamentals justify. The consensus remains bullish: multiple market-data services show analyst-average targets still far above current levels (with a wide high/low target range, reflecting genuine uncertainty about how durable the margin structure is). That spread is the market’s honest admission that it doesn’t know yet whether this cycle is different.

The cycle-is-different argument rests on take-or-pay contract structures and a demand source — AI training and inference — that doesn’t behave like historical PC or smartphone demand. The cycle-is-the-same argument rests on capacity: Micron’s roughly $9.3 billion (¥1.5 trillion) Hiroshima expansion, SK Hynix’s expansion, and Samsung’s quality recovery could all come online roughly simultaneously in 2027–2028. If AI model efficiency improves faster than compute demand grows, the demand side softens just as new supply peaks.

Slight tangent worth noting: Micron signed long-term automotive supply agreements with Qualcomm, Visteon, HARMAN, JOYNEXT, DENSO, Astemo, and Hyundai Mobis (announced July 16). The stock still fell that day. When good news fails to hold a bid, the technical structure is telling you something about where sentiment sits. The 50-day EMA near ~$869 and a deeper support level around ~$776 are what the chart-watchers are watching right now.

Amazon’s ~$200 Billion Question — Due July 30

The connection to Micron is structural. Amazon — a major AI infrastructure buyer — reports Q2 2026 results on July 30. The headline number will be AWS growth. AWS grew 28% year over year in Q1 2026, reaching $37.587 billion at a 37.7% operating margin — its fastest pace in 15 quarters. If Q2 sustains or accelerates that rate, it directly validates the demand thesis that underpins Micron’s bull case. If it softens, both stocks face a recalibration.

Amazon guided Q2 net sales of $194.0 billion to $199.0 billion, with operating income of $20.0 billion to $24.0 billion. Prime Day — which shifted from Q3 into Q2 this year — provides a meaningful revenue comparison tailwind. But Prime Day retail is not what moves AMZN’s multiple. AWS is.

What’s interesting is that Amazon’s custom silicon story may be under-appreciated in the current narrative. In Q1, Amazon said its chips business (inclusive of Graviton, Trainium, and Nitro) topped a $20 billion annual revenue run rate with triple-digit year-over-year growth.

Trainium capacity has also become a concrete part of the capex-to-demand debate: OpenAI and Amazon announced a strategic partnership in which OpenAI committed to consume approximately 2 gigawatts of Trainium capacity through AWS infrastructure. Separately, Anthropic disclosed an agreement with Amazon to secure up to 5 gigawatts of capacity for training and deploying Claude (including nearly 1 GW of Trainium2/Trainium3 capacity coming online by the end of 2026). These are capacity reservations, not just exploratory discussions.

The tension is this: Amazon reported trailing-12-month free cash flow of $1.2 billion versus $25.9 billion a year earlier, driven primarily by higher property and equipment purchases. Amazon also disclosed Q1 2026 cash capital expenditures of $43.2 billion. Amazon’s 52-week range runs from $196.00 to $278.56. As of mid-July it trades roughly in the $250 range — a much shallower pullback than what semiconductors have absorbed. The market is, at least for now, still giving Amazon the benefit of the doubt that this is the 2014–2016 AWS investment cycle rather than a demand overshoot.

AMZN’s 52-week range runs from $196 to $278.56, with the stock currently in the $250–255 range. With over 60 analysts on a buy or strong buy, the sentiment wall is built. What July 30 answers is whether the income statement has started absorbing the upside from the spending.

Technical Structure — Two Different Stories

MU is in a clearly damaged technical setup. The $897.80 support level has been breached. The stock is below its 50-day moving average. RSI readings from multiple services show the stock approaching oversold territory, and positive RSI divergence has been noted at the trendline — which historically precedes a stabilization. The levels to watch are $869 (50-day EMA support), $776 (the next structural downside target), and $715 below that. On the upside, $1,014 is the first meaningful resistance before the post-earnings high near $1,255 becomes relevant again.

AMZN’s structure is more constructive. The stock bounced from its early-July lows and has been holding above the $245–248 range with volume patterns that suggest institutional accumulation rather than distribution. The VWAP from the post-Q1 earnings move sits near $249. A close above $258 would re-establish the bullish bias ahead of the July 30 print. The downside scenario of renewed capex concern sends AMZN back toward the $225–230 range, where it found buyers in early June.

Scenario Framework

Bull Case — MU: AWS July 30 print confirms 28%+ growth and validates AI infrastructure spending. MU’s Q4 $50 billion guidance proves durable. CXMT IPO concerns fade as technology gap proves persistent. HBM4 mass shipments beginning now pull forward the next product cycle. Short-covering from elevated short interest accelerates a move back toward $1,000. Target: $1,000–$1,090 near-term; ~$1,486 consensus within 12 months.

Base Case — MU: Continued overhang from CXMT and export restriction headlines keeps the stock range-bound between $800 and $950. Q4 results in September re-anchor the narrative on fundamentals. The cycle-is-different thesis gets partial credit — memory pricing holds through 2026 but margin deceleration begins confirming in Q1 FY2027. Stock drifts sideways-to-up. Target: $850–$950 range into September earnings.

Bear Case — MU: Export restrictions impacting advanced memory materially reduce addressable revenue. Samsung HBM3E qualification for Nvidia accelerates faster than expected, compressing Micron’s near-term pricing power. Capacity from the Hiroshima expansion and SK Hynix’s expansion arrives earlier than guided, triggering a memory pricing correction in 2027. Stock retests $715 support, and multiple compression back toward 5–6x forward earnings begins. Target: $700–$750.

Bull Case — AMZN: Q2 AWS growth accelerates above 28%. Trainium margin data shows cost displacement flowing through the income statement. Prime Day contribution surprises to the upside. Operating income comes in near the high end of the $20B–$24B range. Stock breaks through $278 all-time high. Target: $285–$310.

Base Case — AMZN: AWS grows 26–28%, in line with Q1 pace. Operating income meets guidance midpoint. Capex narrative remains neutral — no new surprises to the upside. Stock drifts up modestly post-earnings, re-rates gradually as Trainium economics become more visible. Target: $260–$270.

Bear Case — AMZN: AWS growth decelerates below 25%. Capex guidance for H2 2026 comes in above current ~$200B expectation. Free cash flow remains pressured with no timeline given for normalization. Stock sells off to $225–$235 range. Target: $225.

Active Trader Framework

The two trades are linked at the thesis level but structurally independent in the short run. MU is a sentiment trade first — the operational bull case remains intact, but institutional positioning needs a catalyst to recover. The September earnings date is too far out to anchor near-term risk. What matters now is AWS July 30 and any Nvidia commentary that affirms AI accelerator demand. Both would remove the demand question from MU’s bear case without directly resolving the supply competition question.

Traders watching MU should treat $869 as the key level. A close below it on elevated volume with no macro or sector catalyst is a signal the technical damage is deepening, not bottoming. A hold of $869 with declining volume on down days is a different picture. Position sizing matters in a stock with a 52-week range of $1,152 — this is not a range-trading instrument.

For AMZN, the July 30 print is a binary event in the sense that AWS growth above or below 26% will materially reset consensus. The options market into earnings reflects that. Traders who want to express a view on AI infrastructure spending without the binary capex risk embedded in AMZN may find more defined-risk structures in the custom silicon story directly — though that involves names with their own complexities.

The part people tend to skip: the FOMC rate decision lands on July 29 (the day before Amazon reports). If the Fed signals anything that changes the cost-of-capital calculus for $200 billion spending programs, that lands on AMZN’s earnings setup before the print hits. The sequencing matters. Watch July 29 closely.

Micron’s paradox is that a perfect quarter created a valuation that needed no bad news to decline — the absence of certainty about what the next cycle looks like was enough. Amazon’s question is simpler and harder at the same time: can you spend ~$200 billion a year and grow into it? The market has one more major data point coming in 13 days.

For informational and educational purposes only. Not investment advice. Trading involves risk, including loss of principal.