View summary- Akamai–Anthropic deal: 💰 $11.6bn 7-year accelerating CPU agreement announced 24 Sep (potential +$9bn expansion). Anthropic to use Akamai’s distributed AI infra & software for growing CPU workloads. 
🤖 Vera CPUs expected: Nvidia previously flagged Akamai’s Vera deployment plans. Vera targets agentic workloads (code, orchestration, data processing) with high-bandwidth LPDDR5X. 
• Positive read-throughs:
– Server CPU & DRAM demand (agentic AI needs more compute + memory)
– eSSD / NAND (context, vector DB, KV-cache growth; better cost/power vs DRAM/HDD)
– ABF substrates
– Lenovo (major server OEM for Akamai)
• Our Server CPU TAM: 2030 server CPU TAM of $215bn 
• Stocks & Positive Implications :
– CPU: NVDA, AMD, INTC
– Memory: SK Hynix, Micron (DRAM/HBM); Samsung, SK Hynix, MU + SanDisk (NAND)
– Advanced packaging: ASE (Buy), TSMC (Hold), Amkor
– OEM: Lenovo (Buy) #anthropic# #NVDA# #CPU# #DRAM#

Summary of : NAND Pricing Breakout Driven by eSSD ☕️ According to TrendForce, the leading third‑party memory research firm, NAND demand is tracking materially stronger than expected. 📈NAND demand tracking much stronger than expected — driven by eSSD as U.S. CSPs aggressively scale AI infrastructure • WE RAISING NAND blended price outlook meaningfully:
3Q26 / 4Q26 / 1Q27 QoQ: +21% / +17% / +15%
(vs prior +18% / +2% / 0%) • Inventory at U.S. hyperscalers expected to fall further by 4Q26 (from ~14 weeks in July vs ~15-week normal) • Supply remains disciplined: bit growth +15% / +21% / +20% YoY in 2025/26/27 • eSSD demand from U.S. hyperscalers has nearly doubled vs prior projections → demand growth outpacing supply near quarters • enterprise (especially AI-related eSSD) px/demand far outpacing consumer • AI is reshaping the memory hierarchy → new stack: DRAM → SCM → TLC/QLC NAND → HDD • Still bullish DRAM/HBM (SK Hynix, MU) and Now also constructive on NAND names (Samsung, SK Hynix, MU, SanDisk, Phison, etc) #Trendforce# #Memory# #DRAM# #NAND# #MU# #SKHY# #SNDK#

Summary of : NAND Pricing Breakout Driven by eSSD • NAND demand tracking much stronger than expected — driven by eSSD as U.S. CSPs aggressively scale AI infrastructure • Raising NAND blended price outlook meaningfully:
3Q26 / 4Q26 / 1Q27 QoQ: +21% / +17% / +15%
(vs prior +18% / +2% / 0%) • Inventory at U.S. hyperscalers expected to fall further by 4Q26 (from ~14 weeks in July vs ~15-week normal) • Supply remains disciplined: bit growth +15% / +21% / +20% YoY in 2025/26/27 • eSSD demand from U.S. hyperscalers has nearly doubled vs prior projections → demand growth outpacing supply near quarters • enterprise (especially AI-related eSSD) px/demand far outpacing consumer • AI is reshaping the memory hierarchy → new stack: DRAM → SCM → TLC/QLC NAND → HDD • Still bullish DRAM/HBM (SK Hynix, MU) and Now also constructive on NAND names (Samsung, SK Hynix, MU, SanDisk, Phison, etc) #Memory# #DRAM# #NAND# #MU# #SKHY# #SNDK#

Re: CCL Upgrade Debate Theoretically right, but practically the CCL upgrade stays. 1. CPO won’t hit the compute board anytime soon 2. NVSwitch NPO/CPO next year still uses higher-spec CCL (PTFE + M8/9). -> Key thesis is still PTFE 3. Others can be seen in many posts #CCL# #PCB#

Inline with ours on 14A - installed capacity of 6k by 4Q27 with first round of volume input in 1Q28. Current D0 ratio is tracking well. Reiterate our post yesterday ⬇️ •Foundry yields improving:
– 18A (Panther Lake) ~80% in 2Q26
– Larger server CPU (CWF) yields up to ~50% (from ~20% in June) •Capacity reiterated: 40/60/80 KPM (18A) and 0/6/24 KPM (14A) by end-2026/27/28. #INTC#

Apple (AAPL) research summary • iPhone 18 Pro/Pro Max pre-order wait times look lukewarm
• 3-4 weeks for most 256/512GB Pro Max + Burgundy 1TB (US)
• Faster: 1-2 weeks Black/Glacier 1TB, immediate for many Silver 1TB & 2TB + most Pro configs • Limited upgrades + higher pricing (esp. high storage) + Duo competition + weight criticism • iPhone Duo early feedback positive: smooth fold, strong inner display, better iOS 27 multitasking
• Trade-offs: heavier/thicker, visible crease, dual camera only • cutting Duo 2026 build plan 7m → 6m units due to hinge 
• cutting iPhone 18 Pro/Pro Max build plan to 72m on adjustable-aperture camera constraints • Maintain Hold
 #AAPL# #iPhone#

Intel mid-quarter update •Raise TP to $150 on SOTP, citing better-than-expected server CPU (18A) & EMIB yields + foundry client expansion •CPU tailwinds: GPT-6 agentic capabilities expected to boost server CPU demand. •CWF (Clear Wafer Forest): small volume late-2026 → full production 1Q27. Major volume = Diamond Rapids (mid-2027) with stronger SRAM/LPDDR5X + more cores. •NVDA: potential PC cooperation + likely x86 CPU adoption for head-node (Rosa Feynman variant which is with Intel CPU). •Foundry yields improving:
– 18A (Panther Lake) still ~80% in 2Q26
– Larger server CPU (CWF) yields up to ~50% (from ~20% in June) •Capacity reiterated: 40/60/80 KPM (18A) and 0/6/24 KPM (14A) by end-2026/27/28. •Foundry Projects advancing: Terafab cooperation + HBM base die (likely 2029 for 2 major suppliers). •Backend pipeline robust ($1.1bn/$7bn rev 2027E/2028E). •EMIB/substrate yields tracking ahead of market concerns: Japanese suppliers already at ~45% (from 20–25% in 2Q), targeting >50% by end of Sept and ~55% in Jan. @INTC #EMIB# #HBM#

Takeaways from the Sept Wrap-Up report - Cautious Market, Intact AI •July was tough; mild recovery through Aug/1H-Sept still felt challenging. Positioning remains cautious despite robust AI-infrastructure demand. Sentiment is more on worst-case capex & monetization fears. •Worst of the pullback looks behind us: positioning cleaned up, Sept rate hike was expected, and recent tech-CEO comments reinforced AI confidence. •Long-term AI/capex: US CSP capex forecast +85% / +45% / +20% YoY for 2026/27/28. CSP balance sheets (ex-Oracle) stay healthy — net leverage ~0.5x, net debt-to-equity only 18% by 2029. •Neoclouds + Nvidia’s $530bn off-balance-sheet backstop = Nvidia wins if demand holds, manageable if it weakens •GPT-6 is a step-up toward AGI: Unlocks monetizable automation across knowledge work, enterprise agents, DevOps/security, synthetic data/RL. •Scaling + inferencing demand continues even as CEOs talk about “pacing the frontier.” •Recent supply tour + SEMICON 2026: bullish on AI infra demand. Tight supply across wafer fab, advanced packaging, memory, LPS, substrate, DSP, etc. Strong backlogs + multi-year visibility. •Still like: NVDA, CPU, UMC/INTC, DRAM/HBM, thermal, interconnect/optics (Credo, SMTC), PCB/PTFE. •August data: TSMC sales +10% MoM (tracking inline). Neoclouds momentum helps OEMs. Substrate upcycle intact, CCL margin expansion. Aspeed +118% YoY with solid 4Q guide / likely BMC price hike. #AI# #Semis# #NVDA# #MU# #TSMC# #capex#

Memory: Datacenter HBM unlikely to move to 4-Hi 🔑 We share TrendForce’s view that the chance of de-spec to 4-Hi is low, vs worries over 4-Hi which could deliver same bandwidth with fewer dies, cut inference costs & stretch limited DRAM supply • Two key reasons 4-Hi unlikely fly in datacenter AI:
– Per-stack capacity is too small (~16GB for HBM5/HBM4E) → hyperscalers would need more chips, racks & scale-up networking
– Higher volume still increases base-die demand on tight advanced logic nodes (esp. N2 for HBM5) → little to no supply relief • Supply-chain feasibility low: mainstream production is focused on 8-Hi & 12-Hi, vs development cost and low-volume 4-Hi flows • Cost per Gb is higher on 4-Hi vs 8-Hi, with similar total cube cost based on current price quote. • Nvidia’s Rubin Ultra dropping to 8-Hi is well expected (our note on Aug 6) • For Rubin Ultra we expect parallel development of HBM4 8-Hi 192GB / HBM4E 8-Hi 256GB / HBM4E 12-Hi 384GB, with higher chance of HBM4 8-Hi first, then HBM4E 8-Hi as the mainstream SKU (consistent with our earlier AI Thermal report of maintain 2600W) • Reiterate our positive view on the DRAM/HBM sector #HBM# #DRAM# #MU#

Not so research related Very disappointing handling of the severe CX473 delay by @cathaypacific. Important updates were given only verbally to passengers near the front of the gate, while others received nothing — despite a PA system being available. It wasn’t based on status or priority/membership, just where you happened to sit. For a major delay, every passenger deserves the same information at the same time. Communication should be consistent, transparent and fair.

PCB: Content Upside & PTFE Adoption Following our very early PTFE note on June 2, we provide update: • Nvidia Rubin platform a major catalyst for PCB content expansion
GB300 → ~US$440/GPU
VR200 → ~US$750/GPU
VR300 → US$1,000–1,500/GPU • Drivers: higher-end CCL materials, rising layer counts, and potential PTFE adoption into Rubin Ultra • VR200 upgrades
Compute tray: M8 + M6 CCL
Switch tray + new 44L midplane: M8.5 (Low-DK2/HVLP4)
→ Content rises to ~US$750 per GPU • VR300 further upside: M8.5 + M6 hybrid compute + M9 (Q-glass) switch/midplane → ~US$1,000/GPU
Oberon NVL576: hybrid PTFE switch boards (16+16L) under evaluation
If PTFE adopted → content could exceed US$1,500 per GPU • Stock preference
PCB: ZDT (4958 TT Buy) – new position in VR200 compute tray
PTFE CCL: Leading Chinese player primary; Taiflex (8039 TT) as potential second source #PCB# #PTFE# #EMC# #TAIFLEX# #CCL#

Memory: Our Tech Tour Suggests Intact DRAM • Recent supply chain tour + SEMICON 2026 point to stronger DRAM outlook than feared. Turning positive on DRAM sector after prior bearish notes on July 3. • DRAM solid: We believe the US hyperscaler inventories low and likely to be lowered further. Blended DRAM px expected ~10% QoQ in 4Q26 (vs high-teens in 3Q26). • NAND soft: High-teen QoQ pricing in 3Q26, then softening to low single-digit in 4Q26 on weaker mobile demand + elevated inventories. • HBM firm: Constructive 2027 view on tight supply + Chinese GPU demand. HBM4 asking ranging from $2.9-4.5/Gb and we expect the final pricing to be $3.5-4.0/Gb. However, still favor 8-Hi adoption but low probability of 4-Hi GPU variants. Vera SOCAMM de-spec remains, to 64GB by 1Q27. • SEMICON / NVHBM: MTK + Nvidia collab on NVHBM highlighted. Shifts base-die design value toward Nvidia (NVDA) and MTK (2454 TT). cHBM/zHBM progress noted during SEMICON #Memory# #DRAM# #NAND# #SEMICON# #NVDA# #MTK# #MU#

Summary of Credo F1Q27 Results Note 📝 Revenue $479m (+115% YoY, +10% QoQ), broadly in line; EPS $1.20 slightly beat consensus. * F2Q27 guide: Revenue $525–535m; GM 67–69%, both in line. * FY27E growth: Mgmt expects >85% YoY revenue growth, driven by surging AI interconnect demand. * AEC remains the near-term growth engine, with adoption broadening across hyperscalers and neo-cloud customers; 200G/lane and PCIe 6.0 platforms gaining traction. * Longer term, Credo sees AEC/optics approaching a ~50/50 revenue mix. * Optical FY27E revenue maintained at >$600m, with meaningful ramp weighted toward F2H27. * ZeroFlap pluggable transceivers expected to become the largest optical rev contributor, supported by higher ASPs and reliability-focused design. * Higher optical mix creates GM dilution, though absolute profitability remains healthy. * NPO OE, leveraging DustPhotonics silicon photonics, targets sampling in late FY27E; material revenue expected in FY28E. * Maintain $227 TP, based on 20x FY28E EPS; FY27E/FY28E EPS forecast at $6.5/$11.4. #CRDO# #OPTICAL# #AI# #AEC#

Summary of MTK (2454 TT): NVDA’s $3.5bn investment + intact TPU progress 📈 Nvidia to invest $3.5bn in MediaTek via its $3.9bn CB. Alphabet also participating in the CB. * This marks Nvidia’s first investment in a Taiwan-listed company. * Nvidia/MTK expanding collaboration across AI infrastructure, local AI and automotive. * MTK to adopt NVLink Fusion for custom XPU integration. * Collaboration spans NVLink/NVLink-C2C, NVHBM, advanced packaging and SerDes. * MTK/Nvidia aim to provide chip-to-“””rack infrastructure”””, not just silicon. 🐟 Humufish EMIB progress is better than feared. * Japanese substrate suppliers’ yields improved from 20–25% → ~45% on recent testing vehicles. * Yield remains on track toward the first-stage 55–60% target. * Humufish lifecycle volume appears revised up, not cut. * Foe now, We maintain 4.6m Humufish/Triggerfish units in 2028E. * Triggerfish’s large SRAM should help MTK expand into AI inference. We expect MTK to gain TPU share across volume, value and workloads. 🤝Nvidia + Alphabet backing reinforces confidence in MTK’s datacenter trajectory. 📈 Remain positive #NVDA# #GOOG# #TPU# #Mediatek#

Marvell (MRVL) F2Q Results • Stock -7.8% AH on high expectations; Investors await Investor Day in Oct. • F2Q largely inline: $2.74bn rev (+13% QoQ / +37% YoY) • Non-GAAP EPS $0.94 (slight beat) • F3Q guide solid: $3.15bn (+15% QoQ / +50% YoY) • Raised FY27E rev to ~$12bn, FY28E to ~$18bn • Custom AI ASIC: management guiding >100% growth in FY28 • We see path to $16bn+ by FY29 (vs their ~$10bn) • Google multi-year deal covers accelerators, NICs, CXL, near-memory • Another big hyperscaler XPU project progressing well • Connectivity is the real engine behind the $18bn target • 800G still strong, 1.6T DSPs ramping • 51.2T switches expected to double this year • AI clusters need ~10× more bandwidth → big tailwind for 1.6T coherent • Celestial AI deal boosts optical / CPO / NPO capabilities (a big advantage to compete AWS T4) – Buy | TP $290 on 20x FY29E #MRVL# #ASIC# #aws#

Dell (DELL) F2Q Preview • F2Q27E preview soft: $42.2bn rev (below $44.5bn guidance mid) • GM 18.3%, EPS $4.70 (vs cons $4.84) • Expect AI server guidance raise to ~$75bn • That should lift full-year numbers, but market already prices it in • NVL72 shipments cut to 17k / 20k in FY27E / FY28E • Traditional servers still +75%, CSG only +4% in FY27E • Early signs of enterprise & PC demand cooling after strong 1H26 • Competition: Hon Hai winning more SpaceX orders on Rubin • Positives: Tier-2 Neoclouds (IREN, NScale, Crusoe) building out • Vera CPU racks shipping to Oracle + SpaceX • Stock still looks expensive at ~20x FY28E EPS. Maintain HOLD #DELL# #Rubin# #SpaceX#