MojoTrick
@MojoTricks
Curating the latest in AI, semiconductors, EVs & global tech breakthroughs | Not affiliated with any brand
61 Following 27.3K Followers
Intel annual free cash flow (2005–2025):
Peak $21.4B (2020) → –$15.7B (2024) → improving to –$4.9B (2025).
Massive CapEx for IDM 2.0 crushed FCF.
Sandisk Data Center revenue just went from $40M (Dec ’23) to $4.8B (Oct ’26).
That’s not a cycle. That’s a structural shift.
Demand > Supply. Multi-year commitments locked in. Traditional NAND volatility is getting rewritten in real time.
$SNDK
Show more
Operating margins of the big three memory makers continue to climb.
Micron, SK hynix and Samsung DS all saw sharp expansion from 4Q25 to 3Q26.
Samsung DS still trails but has recovered strongly.
Note: Micron’s fiscal quarters differ (3Q26 = 4QFY26)
#Samsung#
Show more
Nomura estimates: Global data center power by applicationAI goes from 10GW (17%) in 2025 → 40GW (35%) by 2028Heavy agents become the biggest AI power user (15GW), ahead of training (10GW)
Show more
Memory stocks trading at just 2–3x earnings.The market is already pricing in a 2027 downcycle that still hasn’t shown up in the 2027: current 3.2x vs DCF-implied 9.1x.
Show more
Micron just laid out its full memory roadmap through 2030.
• DDR6: >2× bandwidth vs DDR5
• LPDDR6: >35% better power efficiency (1δ node)
• HBM5 (2028) → HBM5E (2029) → HBM6 (2030)
• vHBM with hybrid bonding starting 2028
• G11 NAND: >400 layers, 2× density vs G9, ramp in 2H27
Vertical stacking is coming. The memory wall is real.
#Micron# #HBM# #DDR6# #NAND# #AI# #Semiconductors# $MU
Show more
Micron just laid out its AI memory roadmap through 2030.
Token efficiency = Bandwidth + Energy Efficiency + Capacity
• HBM (JEDEC + Custom) → HBM4 to HBM6
• vHBM (Vertical 3D HBM) starts 2028
• vEIM (Vertical Edge Inference Memory) targets 2TB/s & <1pJ/bit
Vertical stacking is coming. Memory wall is real.
#HBM# #Micron# #AI# #Semiconductors# #vHBM#
Show more
Samsung’s HBM comeback is real.
Revenue share
Q2 2025 → Q2 2026
SK hynix: 64% → 50%
Samsung: 15% → 33%
Micron: 21% → 18%
Doubled in 12 months. Still trailing SK hynix, but the gap is closing fast as HBM4 ramps.
Show more
Samsung & SK hynix upgrading China NAND fabs past 200 layers as YMTC closes in with 294L. AI demand vs. export controls - the race is on. #NAND# #Semiconductors#
NAND spot: still weak on the module side.
512Gb TLC ▼1.28% | eMMC ▼1.25% | MicroSD ▼3.91% | MLC ▲0.92%
High inventories, sluggish trading. TrendForce, Oct 19.
#NAND# #MemoryPrices# #AIStorage#
Show more
AI accelerator market is projected to hit $1.46T by 2030, with Nvidia alone taking ~$1T (~69% share) according to BofA estimates.
Stacked growth from 2024–2030 shows the scale of the AI infrastructure buildout.
#NVDA# #AI# #Semiconductors# #BofA#
Show more
Memory stocks took a heavy hit after CXMT’s 4F² DRAM news - major suppliers down 11-26% in days.
Here’s the Goldman Sachs breakdown, animated bar by bar.
$MU $SKHY $SNDK
Intel ASIC services → ~$2B run rate (3x YoY). Path to $4B. Custom AI silicon for hyperscalers is the real story. $INTC
AI infrastructure is still extremely tight.
TrendForce data (Oct 19, 2026) shows hyperscalers locking HDD supply into 2028 yet still expecting shortages. Lead times remain severely elevated - HDD at 50 weeks vs 16 weeks balanced, ABF at 48–56 weeks.
Here’s how current lead times stack up across key components:
HDD • ABF • DRAM • MLCC • CPU • NAND/eSSD • GPU
The squeeze isn’t easing anytime soon.
Show more
Sandisk FY2026: $20.2B revenue (+175%)
Datacenter: $5.2B (+437%)
Edge: $12.2B
Consumer: $2.9B
The AI storage wave is here.
$SNDK #Sandisk#
AI data centers are about to explode 📈
Bottom-up analysis shows ~212GW of incremental capacity needed from 2026–2030.
Annual deployment:
• 2025: 13.0 GW
• 2030E: 55.7 GW
Nvidia still dominates, but AMD, Google TPU, AWS Trainium, Microsoft Maia, Meta MTIA & OpenAI are all scaling fast.
Show more
Goldman Sachs just raised its Global Server Value TAM projections:
📈 +31% in 2027E ($1.11T)
🚀 +45% in 2028E ($1.60T)
General servers remain flat/steady while AI servers swallow the entire market share.
#Semiconductors# #AI# #DataCenters#
Show more
$NVDA's outperformance is turning the rest of the Magnificent 7 into defensive plays
Most people mix these two up.
DRAM ≠ storage
NAND ≠ RAM
One is temporary working memory.
The other is permanent storage.
Quick visual breakdown:
Samsung & SK Hynix currently trading at just 2-3x P/E on 2027-28E EPS. P/B also near 1x - levels typically seen at cycle troughs.
Micron still trading at a premium.
#Memory# #Semiconductors# #Samsung# #SKHynix# #micronesian#
Show more
