Newsletter · 2026-07-04 · Ashutosh Agarwal

Micron's Blowout Quarter and the Memory Supercycle Debate - Semiconductor Podcast Briefing - Week of July 4, 2026

Micron's record quarter (84.9% gross margin, roughly $100B of contracted revenue through 2030) reset the debate over whether memory is a cyclical spike or a structural regime change, while Qualcomm entered data-center silicon and Apple lobbied Washington to buy blacklisted Chinese memory. Our synthesis of the semiconductor podcast tape for the week ending July 4, 2026.

Semiconductor Podcast Briefing

Week of July 4, 2026: Micron's Blowout Quarter and the Memory Supercycle Debate

Week ending July 4, 2026 (coverage window ~June 25 – July 3, anchored by Micron's June 24/25 print)

This week's podcast conversation was dominated by one theme: memory. Micron's blowout quarter reset the debate over whether this is another cyclical spike or a structural regime change, and the second-order effects rippled everywhere: Apple raising hardware prices, hyperscalers accelerating custom-silicon programs to route around both NVIDIA and the memory "mafia," and Apple lobbying Washington to buy from blacklisted Chinese memory makers. Semis closed out their best quarter ever.

TL;DR, Five Things That Mattered This Week

1. Micron's quarter was "one for the ages." Gross margin hit a record 84.9% (up ~10+ points), with $25B operating cash flow and $18B free cash flow in the quarter; revenue was over $41B with Q4 (Aug-year) guidance of ~$50B and implied operating income over $41B. The stock was up ~300% YTD and ~800% over the trailing twelve months at a ~$1.3T market cap. Management guided to memory shortages persisting into 2028 with "no line of sight" to supply catching demand. (Squawk on the Street, Bloomberg Intelligence)
2. Memory pricing is vertical. DRAM ASPs cited up ~60%, NAND up ~80% quarter-on-quarter; other commentators put DRAM/NAND 300–400% above year-ago levels. Microsoft said memory prices are up 2.5x and expected to double again by fall 2027. Apple raised Mac/iPad prices $100–$300 (17–25%) mid-cycle. (Halftime Report, Squawk on the Street, Rob Black Show)
3. "Take-or-pay" is reshaping the memory cycle, but the bull/bear debate is unresolved. Micron has signed 16 strategic customer agreements running 2026–2030, ~$100B contracted revenue, expected to reach ~40% of revenue, with ~$22B of customer cash deposits landing on the balance sheet. Bulls call memory an "irreplaceable resource"; bears note it's all price, not volume, and that 85% margins invite capacity and customer workarounds. (Chip Stock Investor, Bloomberg Intelligence)
4. Qualcomm's data-center debut was the week's other big story. At its investor day, Qualcomm raised its FY2029 non-handset revenue target to $40B (from $22B), including $15B from data center (with ~$5B next year), under a new "Dragonfly" brand, with custom-chip engagements starting in 2027 for two hyperscalers (Meta, Microsoft). (Schwab Network)
5. Apple is lobbying Washington to use blacklisted Chinese memory. Apple is seeking flexibility to buy DRAM from CXMT (and NAND from YMTC), which can undercut incumbents by up to 30%. YMTC (~13% global NAND) is on the Commerce entity list; CXMT (~8% DRAM) is not yet, and industry is lobbying to keep it off ahead of a mainland IPO. China hawks including Sen. Tom Cotton call it a "grave mistake." (CNBC Fast Money, Squawk on the Street)

1. AI Chip Demand & Hyperscaler Capex (NVDA, AMD, AVGO, MRVL)

The demand narrative held, but the week's texture was about who captures the spend. Bloomberg Intelligence semiconductor analyst Jake Silverman (June 25) framed the memory pull as capex-driven: "We're seeing hyperscale CapEx continue to grow at an incredibly high rate and memory is becoming an increasingly large portion of that... We're shifting from an AI training environment to AI inferencing, and that's just increasing the amount of memory you need in these data centers." (Bloomberg Intelligence)

The picks-and-shovels question resurfaced. On CNBC's Squawk on the Street (June 25), Mike Santoli captured the crowding: "every type of strategy that seems to make sense right now lands you into owning semis... earnings momentum, earnings revision trends, price momentum, growth at a reasonable price. All those strategies funnel you into the same [trade]." He also flagged concentration risk, semis are now above 20% of market cap and Micron's beta was running 3.1 ("three times as volatile as the S&P 500"). (Squawk on the Street)

Custom silicon accelerated as a hedge against NVIDIA. Multiple episodes covered OpenAI's "Jalapeño" inference chip (built with Broadcom) and reporting that Anthropic is in talks with Samsung for a custom chip, both framed explicitly as reducing NVIDIA dependence amid rationed GPU allocations (AI Update, Equity). On Sequoia's Training Data (June 30), Dylan Patel of SemiAnalysis noted Google is deploying "hundreds of billions of dollars per year in their own ASICs," running multiple distinct TPU architectures with Broadcom and MediaTek, while cautioning that TPUs underperform on some models, and that hardware-software co-optimization, not any single chip, drives the largest efficiency gains (Training Data).

Contrarian / cautionary notes. David Bahnsen (The Dividend Cafe, July 3) observed that hyperscaler capex as a share of cash flow has more than doubled, financed increasingly by debt and equity, and that NVIDIA, "the mother of all pick-and-shovel companies", was actually down from its late-October level even as memory suppliers soared, underscoring how unevenly the AI trade is rewarding participants (The Dividend Cafe). Palantir CEO Alex Karp (Squawk Pod, July 1) said Palantir has aligned with NVIDIA around customer control of compute/models/data, while enterprises grow wary of frontier-model token pricing (Squawk Pod).

2. Memory Pricing (HBM / DRAM / NAND, MU, SK Hynix, Samsung)

This was the week's center of gravity. CEO Sanjay Mehrotra, quoted from Micron's call on Squawk on the Street (June 25): "With respect to supply, our customers are recognizing that supply shortages in memory and storage will take considerable time to improve. Even as we expect industry supply to improve gradually in 2028, we currently do not have line of sight as to when memory supply will be able to catch up with increasing demand." (Squawk on the Street)

The price magnitudes cited across shows:

• DRAM ASPs +60%, NAND +80% (Halftime Report, June 25) (Halftime Report)
• Micron revenue +74% QoQ, +346% YoY; DRAM prices up 3–4x and storage tripled over the past year as capacity shifts to AI (The Canadian Investor, June 27) (The Canadian Investor)
• Microsoft: memory up 2.5x, expected to double again by fall 2027 (Squawk on the Street, June 26) (Squawk on the Street)
• DRAM/NAND 300–400% above year-ago, forcing Apple price hikes of 17–25% (Rob Black Show, June 25) (Rob Black Show)
• Prof G (June 30): DRAM up nearly 100% in Q1, another 60% jump expected, referencing Samsung, SK Hynix and CXMT (The Prof G Pod)

The structural bull case, "irreplaceable resource." On CNBC's Halftime Report (June 25), one panelist described a "paradigm shift... fundamental validation that this is not a cyclical commodity moment for Micron. This is something more powerful... This is an irreplaceable resource... the speed of AI accelerators basically get frozen without the memory component." Unlike oil, the argument goes, there is no near-term price elasticity, and the 16 SCAs "going out to 2029 to secure pricing and revenue" cement it. (Halftime Report)

The bear / cyclical rebuttal. The same program aired the counter: "Micron is achieving all of this on price, not on higher volumes... there comes a point at which the pricing power becomes so extreme that the people who use the product find workarounds", invoking the DeepSeek episode as precedent for efficiency shocks. Mike Santoli on Squawk on the Street (June 25) was blunt: "normally capitalism doesn't say you guys get 85% margins indefinitely... It's still a cyclical business," adding that Micron has traded down after five of the last six earnings reports (down 20% and eight straight days after the prior print). (Halftime Report, Squawk on the Street)

On the Korean names and capacity response. John J. Hardy (Saxo Market Call, June 26) noted Micron rose 15.7% and the SOX +3.59%, but Samsung and SK Hynix sold off on Samsung's announced multi-hundred-billion capex program (cited at ~$650B over 10 years), the market pricing ROIC skepticism into the cyclical memory build (Saxo Market Call). Jake Silverman added that SK Hynix has a stronger HBM demand base than Micron given its capacity and NVIDIA history, and reiterated: "I still think that memory is cyclical. It's just that we might see some structural changes in the way that the pricing volatility changes." (Bloomberg Intelligence)

3. The "Take-or-Pay" Contract Structure, A Possible Regime Change

The most analytically substantive segment came from Nicholas and Kasey Rossolillo on Chip Stock Investor (June 30), who dissected Micron's strategic customer agreements (SCAs):

• What they are: take-or-pay deals, "You either take the agreed-upon product, or if you decide you want to pull the rug out from under Micron, you still have to pay" a pre-agreed minimum (a cancellation charge).
• Why they exist: to prevent a repeat of 2022–23, when industrial/auto customers double- and triple-booked, then canceled en masse when the cycle flipped, leaving Micron holding the bag.
• Scale: 16 SCAs, five-year terms (2026–2030); hyperscalers (and NVIDIA on their behalf) get ~5-year terms, automotive ~3 years. Some have ceiling and floor prices on existing products; some are fixed-price. At current market prices, SCAs are expected to be ~40% of revenue. Micron will book ~$22B in customer cash deposits as a balance-sheet liability over the coming quarters.
• The caveat: the hosts argue current 80–90% gross / 70%+ operating margins are unsustainably high and will likely moderate by 2028–2029 as Samsung and SK Hynix bring on supply, precisely because those prices incentivize competitors to add capacity. (Chip Stock Investor)

Bloomberg Intelligence framed the same uncertainty around the uncontracted book: Micron has ~25% of revenue contracted, targeting ~50%, "Now that leaves another 50% that isn't potentially going to get contracted. So when we start to see a downturn... what happens to the rest of the business?" (Bloomberg Intelligence)

4. Semiconductor Capital Equipment / WFE (ASML, AMAT, LRCX, KLAC)

Equipment names had a strong week but drew valuation caution. On Chip Stock Investor (June 25), the hosts noted April–May 2026 was "the biggest, fastest run-up ever" for semis, and that the "Fab 5" (ASML, Applied Materials, LAM Research, KLA) are "now sitting at pretty high valuations on the expectation that this is going to be a pretty epic growth cycle that lasts beyond just this year and 2027," adding a note of "deserved caution" on those names (they own all four) (Chip Stock Investor).

IBD's Stock Market Today (June 29) flagged relative-strength leadership: KLA up ~12% at new highs, and ASML holding its 21-day line on strong equipment demand, characterizing equipment makers as a distinct sub-sector from designers (NVIDIA) and foundries (TSMC) (Stock Market Today With IBD). A separate IBD episode (July 1) highlighted AMAT, Onto Innovation and KLA as the strongest chip-equipment performers over the trailing six months (Stock Market Today With IBD). A dividend-focused show separately flagged AMAT trading at ~59x earnings / ~83x free cash flow as "extremely elevated and vulnerable to compression" despite strong fundamentals (Dividend Investing with Longacres Finance).

No dedicated ASML- or LAM-specific earnings episode surfaced this week.

5. Foundry / Manufacturing Dynamics (TSM, INTC, GFS)

Coverage here was lighter but pointed. On Squawk on the Street (June 26), the desk discussed a NYT piece arguing the U.S. is "more reliant on TSMC than ever" despite the CHIPS Act, tariffs and Trump-administration equity stakes. Notably, TSMC did not fall on reporting that Apple could use Intel as a fab for its M-series chips, because "Apple's really lost its pricing power... TSMC gets better margins from the hyperscalers because these wafers only have limited supply," so foundry capacity is prioritized to higher-paying AI customers over Apple (Squawk on the Street).

Bloomberg's Big Take Asia (June 30, Charlotte Yang and Shuli Ren) characterized TSMC as the world's largest, highly profitable foundry with a strong tech edge, but warned Asian AI valuations are stretched, Samsung and SK Hynix "trading like meme stocks" despite being trillion-dollar companies, with tight supply/demand persisting through 2027 (Big Take Asia).

6. Analog / Auto / Industrial Semis (TXN, ADI, MCHP, ON, NXPI, STM)

Thin coverage, but two threads worth noting:

• ON Semiconductor CEO gave an exclusive on Squawk on the Street (June 26) amid the "more tech pain" tape (Squawk on the Street).
• DRAM shortage is now hitting autos "especially hard", Automotive News Daily Drive (July 2) covered how the memory crunch is bleeding into vehicle production (Automotive News Daily Drive). Qualcomm's Nakul Duggal discussed the auto/edge-to-data-center silicon roadmap on multiple mobility shows (Automotive News Daily Drive).

The classic analog names (TXN, ADI, MCHP, NXPI, STM) did not get dedicated podcast treatment this week, a coverage gap to watch as their late-July earnings approach.

7. China / Export Controls & China Indigenization (SMIC, CXMT, YMTC, Huawei)

The Apple–China memory story tied topics 6 and 10 together. On CNBC's Fast Money (June 29), Shahzad Qazi (COO, China Beige Book) and CNBC's Mackenzie Cigalos laid out the mechanics: Apple is "again lobbying the administration for more flexibility to work with Chinese memory suppliers as it deals with one of the worst component crises in its history," with Chinese makers able to undercut incumbents by up to 30%. The nuance:

• YMTC (NAND) has been on the Commerce entity list since 2022 (the list with "a lot more teeth"), requiring approval to share blueprints.
• CXMT (DRAM) is not yet on the entity list; industry is lobbying to keep it off, and it's heading for what could be mainland China's largest IPO in years.
• Counterpoint data: YMTC ~13% of global NAND, CXMT ~8% of DRAM, both gaining share fast.
• Qazi's strategic warning: allowing Chinese memory in would risk "long-term US reliance on China similar to rare earths and pharmaceuticals." Sen. Tom Cotton called blacklisted-chip use a "grave mistake for Apple, risking shareholder value, privacy, and national security." (CNBC Fast Money, Squawk on the Street)

Bloomberg Intelligence (July 2) added that national-security hawks including the House Foreign Relations Committee head oppose the arrangement, and the outcome may hinge on a Trump decision (Bloomberg Intelligence). On Real Vision's Macro Mondays (June 29), the view was that Chinese competitor CXMT is closing the yield gap with Micron and SK Hynix but "hasn't reached parity yet" (Real Vision).

Adjacent supply-chain risk: John Feneck (Palisades Gold Radio, June 25) flagged that China produces ~80% of global tungsten and that Japan shuttered ~25% of global tungsten-hexafluoride supply (a semiconductor precursor) as of January 2026, a lesser-watched input bottleneck (Palisades Gold Radio).

8. Cyclicality / Peak-Trough Debate

Beyond the Micron "irreplaceable resource vs. still cyclical" debate above, the cleanest cyclicality framing came from Closing Bell Overtime (June 30): chip stocks had their best quarter ever (SOX up ~95%) while big tech lagged, with Paul Hickey cautioning that semis "may be extended after such a rapid run" (Closing Bell). Bloomberg Businessweek's "Chip Stocks Put Up Their Best Quarter Ever Despite Wild Swings" echoed the volatility (Bloomberg Businessweek). Bull-side, Jordi Visser (The Pomp Podcast, June 27) argued memory demand from AI agents, humanoids and AVs keeps the shortage running to 2028 and the AI trade "in early innings" (The Pomp Podcast).

Earnings & Investor-Day Reactions

What I'm Watching Next Week

1. The Trump-administration decision on Apple / CXMT / entity list. Whether CXMT gets added to the Commerce entity list (or stays off ahead of its IPO) is the swing factor for the memory "mafia" (MU, Samsung, SK Hynix). A carve-out allowing Chinese memory would likely trigger a memory-name sell-off; a hard "no" keeps the squeeze intact (CNBC Fast Money).
2. Q2 earnings season kickoff for the equipment complex and analog names. ASML (mid-July) sets the WFE tone; TXN, ON, NXPI later in July will test whether the auto/industrial DRAM shortage is helping or hurting analog margins. Watch for confirmation of "Fab 5" valuation concerns.
3. Any iPhone pricing announcement, flagged as the "next shoe" after Mac/iPad hikes; a 20% iPhone price move would validate the AI-driven-inflation thesis and pressure Apple volumes.
4. Micron follow-through. Given the stock's history of fading after earnings (down after 5 of last 6 prints), watch whether the "irreplaceable resource" re-rating holds or reverts to cyclical trading.
5. Semi-specific M&A. No chip-sector M&A surfaced in podcasts this week (the deal chatter, Rocket Lab/Iridium, Fox/Roku, Tesla/SpaceX, was all non-semi). With Goldman flagging an M&A boom, watch for consolidation signals among smaller equipment/analog names.