Micron Technology
-
SK Hynix Q2 2026 Earnings Summary (Demand & Supply)SK Hynix reported record Q2 results with revenue of ~$54.1 billion (+51% QoQ) and operating profit of ~$41.6 billion (+61% QoQ), driven by strong AI demand and tight supply.
Demand: Robust and structural. AI infrastructure investments continue, broadening needs for HBM, server DRAM (agentic AI), and enterprise SSDs. The company expects DRAM demand to grow mid-20% YoY and NAND high-teens % YoY in 2026. Momentum is supported by AI service revenues, with no signs of slowdown.
Supply: Remains constrained. Customer demand exceeds capacity, especially for advanced products. Q3 guidance: DRAM bit shipments +~10% QoQ; NAND +low-single-digit % QoQ. Supply-demand balance unlikely to ease soon due to process complexity and long fab lead times.
Finalised LTAs with ~10 key customers for stability. Raising 2026 CapEx to high-KRW 40T range (~$28–32 billion) to accelerate capacity (HBM4 ramp, new facilities). Timely supply is a core competitive advantage.
-
Recent reports that the US government may investigate Chinese memory maker CXMT have significantly reduced the likelihood of Apple receiving approval to use its DRAM chips. US lawmakers are reportedly concerned about CXMT's alleged links to the Chinese military and broader national security implications, placing the company under increased political scrutiny.
Apple has explored sourcing DRAM from CXMT, particularly for iPhones sold in China, to reduce costs and diversify its supply chain. However, any approval would now face a far more difficult political environment.
The Trump administration would need to balance Apple's commercial interests against concerns over technology security, supply chain resilience and the strategic importance of supporting US semiconductor manufacturers.
Several outcomes remain possible. The administration could reject Apple's request entirely, permit CXMT memory only in devices sold exclusively within China, or impose strict conditions that severely limit its use. Given the current climate, broad approval appears considerably less likely than it did only weeks ago.
For existing memory suppliers such as Micron, Samsung and SK hynix, these developments are potentially positive. If Apple is unable to source meaningful volumes from CXMT, demand is more likely to remain with established suppliers. While the situation remains fluid, the latest developments suggest any significant disruption to the global DRAM market has become less probable in the near term.
-
It’s largely irrelevant even if they get approval. Wafers are scarce. That’s what the hard of thinking don’t get. You’ve got a lot of retail ‘investors’ who are oblivious to the facts. Cxmt control 7% of the ‘dumb’ dram commodity segment. If micron see any increase in supply in that area they’ll simply deploy those wafers elsewhere. Cxmt can not produce ai/dc memory and won’t for ‘years’ and when/if they do they won’t get 1 chip into a western dc. Obvious why but it creates a nice fantasy story for people like Burry to induce weak hands to sell.
Sanjay will prove them wrong, again, soon. All that matters is earnings. Which are growing very rapidly and are very durable.
-
Counterpoint Research’s Q2 2026 DRAM figures mark a net big win for Micron.Samsung led with 39% revenue share, while SK Hynix fell to 26% — just one percentage point ahead of Micron at 25%.
This closes a previously wide gap; SK Hynix has lost share for four straight quarters after peaking at 39% a year earlier.Micron’s DRAM revenue surged roughly fivefold since Q2 2025, far outpacing relative share trends. SK Hynix still grew revenue a robust 214% year-on-year, yet underperformed on market share.
The shift stems from HBM3E prices falling through renegotiations rather than rising, plus HBM4 launch delays. SK Hynix, the heaviest HBM player, signed long-term agreements earliest and thus locked in lower fixed prices before conventional DRAM prices climbed sharply.
Those higher conventional prices favoured Samsung and supported Micron’s gains.
Separately, Micron is executing strongly on HBM4 itself, with volume shipments already under way, the ramp progressing roughly twice as fast as its HBM3E transition, improving yields, and full 2026 HBM capacity sold out under long-term contracts. No material issues have been reported.
For Micron this combination is clearly a stronger profit driver than for SK Hynix. The American firm is now positioned to challenge or overtake second place, depending on capacity and future contract pricing.
-
I should also reiterate ALL HBM is as of todays news, SOLD OUT through 2027 also.
TrendForce: DRAM Supply Shortage to Persist Through 2027… NVIDIA Considering Downgrading Rubin Ultra's HBM Configuration
- According to TrendForce's latest memory industry research, the DRAM supply shortage is expected to continue through 2027.
Note-Nvidia decision to use less HBM is not a negative, although short sellers will try and spin it as such. Nvidia has prioritised increased GPU sales-more racks and there is simply not enough HBM available.
This remains positive for Micron.
The TrendForce view confirms a structural DRAM/HBM shortage lasting through 2027. Even with NVIDIA reviewing lower-stack options (HBM4e 8hi, HBM4 12hi or 8hi) for Rubin Ultra instead of the original 12hi HBM4e baseline, the core constraint is unchanged: wafer allocation limits and demand still outstripping supply.
HBM bit shipments are projected to rise 50–60% year-on-year in 2027, yet this will not match demand growth. Suppliers therefore retain strong pricing power, with substantial HBM price increases widely expected. Lower stack heights may allow more GPUs to ship, but each GPU still requires scarce HBM. Micron, as a major HBM producer, benefits from sustained high average selling prices and elevated margins rather than volume dilution from oversupply. AI chipmakers face both limited availability and rising costs, reinforcing the suppliers’ leverage.
-
Turns out China’s CXMT playing hardball on DRAM prices is actually good news for Samsung, SK Hynix (and by extension Micron) – not the threat people expected.Apple was eyeing CXMT as a cheap alternative source for mobile DRAM like LPDDR5X to cut costs on the next iPhones and gadgets.
But CXMT flat-out refused further discounts and is sticking to prices the same as or higher than Samsung and SK. Apple, usually the big buyer that can force prices down, just got stonewalled.Why? Chinese giants like Huawei and Xiaomi have already snapped up most of CXMT’s output with expensive long-term deals.
They’re pushing hard for domestic supply because of the ongoing US sanctions, so CXMT doesn’t need Apple’s business at a cut-price rate.That old Apple trick of dangling cheap Chinese parts to squeeze better deals out of the big suppliers has stopped working.
At the same time the major memory makers have shifted loads of capacity over to high-value AI stuff like HBM. Commodity DRAM (DDR5, LPDDR5X) is tighter, prices are climbing, and the usual cheap workaround has vanished.CXMT soaking up the China-side commodity demand means Samsung and SK are off the hook for flooding the market with low-margin chips. They’re free to pour everything into premium AI memory – HBM4, LPCAMM2, enterprise SSDs and the rest.
With commodity prices held up by CXMT and the Big Three owning the high-end, their pricing power and margins look set to get even stronger in the second half.
-
I don't see Apple sacrificing much-more likely the consumer takes the brunt if they want the new model, maybe even the old model.
See below. Think about that. 34% of the flagship iPhone cost is 'memory' and expected to be 42% next year. Crazy.
What is happening is Wafers are being allocated to HBM and DC memory types leaving commodity DRAM scarce also.
Musk said over the weekend that supply of memory will increase 20-30% in 2027 but demand will increase 200%. It doesn't matter if that is a gross exaggeration. Demand outstripping supply = prices stay up.
In other news CXMT turned out to be the Emperors new clothes. After hanging an 'open for business' sign out front it turns out the shop is empty. They have no memory to sell to the West, not that any sales would have been approved. But it bolstered their stock price. They are now telling the market they'll build more fabs. Nice story, we will see what they achieve in about 4 years-that is how long it takes. And given they are currently on the 1Z node (a dinosaur), I think this 'threat' has been put to bed-
showKeybanc fireside chat with Micron today. I''ll be ears
s -
I think the sector will remain volatile and range bound until we get some concrete info from senior management. The bears won't capitulate without a fight-no doubts, why, because many are paid to generate FUD. As we have said before, the thesis is not based on ever increasing price rises. 25% bit growth is a given and that'll do imo. I don't see prices dropping-basic economic theory says so.
-
Micron gave a fireside chat today at Keybancs Tech conference, where senior executives spoke about what is happening. Brilliant commentary.
This is far more bullish than the market appears to appreciate.
The latest management comments from Micron are, in my view, exceptionally bullish and point to a significant rerating of the shares.
The key message is that demand is not merely strong — it is running substantially ahead of supply, and the imbalance is getting worse. Management indicated that Micron cannot currently meet even half of customer requirements, versus the previously stated 50–67%. That is an extraordinary statement given the scale of capacity investment already underway.More importantly, management said demand has increased since the last earnings report and that supply will be tighter in 2027 than in 2026. That completely undermines the conventional argument that additional industry capacity will soon normalise the market.
Customers are also demonstrating how seriously they view the shortage. Some are signing supply commitments extending all the way through 2030, while management says customers remain largely insensitive to memory pricing. That combination is hugely significant: customers are effectively prioritising guaranteed supply over price.
And this is already translating into extraordinary economics. Micron achieved an 86% gross margin last quarter, with management describing financial performance as being at “extraordinary levels” and robust (meaning they are probably rising further).This is no longer simply a cyclical memory recovery story. It increasingly looks like a structurally constrained market in which AI demand is overwhelming available supply, customers are locking up capacity years ahead, and Micron has substantial pricing power.
At these levels, the market is still valuing Micron too much like a traditional cyclical memory company.
If these trends persist, MU needs to rerate to reflect structurally higher earnings, margins and visibility. -
you can't see what you don't want to see
Shipping 'now'
The guide was $50B revenue this quarter. I think their gross margin will be $50B
️)
