@sssjeffpu

My 2 cents: First of all, excess capacity is a “if”. Nevertheless, it is not a new news, as Meta is likely responding to investor pressure by leasing part of its compute capacity, if any, as a way to help ROE amid high rental px. Please note that the transformation into Neocloud usually leads to higher capex, and we mentioned in our earlier report of “SpaceX’s Move to Neocloud on June 22” Recent developments suggest that Meta remains compute-constrained. News reported that Google restricted Meta’s access to Gemini compute due to high demand. Meanwhile, Meta recently signed a new $21 billion AI compute agreement with CoreWeave for AI capacity. #Meta #NVDA

Good comments👍 From NVDA perspective, due to the limited area of compute tray, it is reasonable to chose 2-die instead of 4-die, as whatever the package is, a compute tray can only do 8 dies. As said, due to different design, MI550, assuming no further change, is expected to be 4-die

Abstract of MediaTek (2454 TT) Triggerfish Implications: 🚀 Triggerfish Implications: 1Higher ASPs — ~$18k vs Humufish $13k or Zebrafish $5.5k 2v9 Mix — Triggerfish to account for 50% of v9 (incl. Humufish) 3Content — 4x companion compute cores with large SRAM (~250MB x4), one 5nm simulator, 4x I/O cores 4Training → Inference — Large SRAM + simulator CPU enables MTK to extend from training (v8t) into inferencing 📈 vs. existing competitor in TPU: •Rumored Bladerunner (2nm) hasn’t progressed as a project •Only one new TPU SKU (Whalefish/Sunfish x2) expected in 2028, leading to disadvantages in process node and HBM → limited volume potential 📈 MTK’s positioning in TPU •v8: MTK’s v8t to start volume earlier than that of v8i •v9: MTK’s Humufish/Triggerfish is by now the only 2nm project •v10: MTK’s Icefish to be solid, while others uncertain 📈 Overall, MTK share price underperformed in recent weeks despite sell-side’s TP raise. Nevertheless, we believe the Street underestimates Triggerfish and now raise TP to NT$6,420. #mediatek #TPU #GOOG #MTK

Abstract- Dell Down from Buy to Hold: •Valuation stretched: +200% rally since earnings (+200% since our U/G in Oct), >20x FY28E EPS suggests limited upside. •SMCI (U/G on June 22) expected to win more NVL72s for SpaceX’s 2027 GW buildout. •ODM-Direct risk: SpaceX & CoreWeave evaluating direct deals with Quanta/Hon Hai/Pegatron. •PC moderating: Weakness in ODM builds, sequential shipment decline expected in F3Q. #DELL #SMCI

Abstract of “SpaceX Move into Neocloud” •Major Transformation: Anthropic + Google deals signal ~$26bn ARR potential vs. SpaceX AI revenue seen at $3.2bn in 2025. Expect massive compute buildout: 1GW (2026) → 4.5GW (2027) → 9.5GW (2028). •Tailwinds for Tech Supply Chain in 2027-2028 driven by aggressive compute capacity expansion. •GPU & ASIC Winners: ◦Primarily Nvidia (NVDA): ~2.2m / ~2.5m Blackwell-equivalent GPUs in 2027/2028. ◦MediaTek (2454 TT) expected to win first datacenter ASIC project in 4Q28 (said in our June report), leveraging Intel process expertise. ◦Intel (INTC): Key capacity partner of TeraFab via 14A Arizona. •New Buy: Credo (CRDO): Benefits from SpaceX compute ramp + shift to 200G/lane with ZeroFlap AEC edge. Stronger optics revenue expected in FY27/28. Bizlink (3665 TT) also gains. •📈 SMCI Upgrade to Buy: Beyond recent GB300 Colossus 2 orders, expect more big AI server wins from SpaceX starting F2Q27. Stabilized margins + better risk/reward. #spaceX #nvda #crdo #smci #mediatek #intel