🎁Nvidia Earnings Week – Trade & Win!（SGD Cash、Task coins...) ✨✨

$NVIDIA(NVDA.US)is now a full-stack AI infrastructure platform: chips → systems → networks → CUDA → AI factories, plus upstream co-investment with third parties to lock in AI factory "Longevity Preservation Systems" (LPS).

Related ETFs:$NVDA 2X Long ETF(NVDL.US)$Direxion Daily NVDA Bear 1X ETF(NVDD.US)

1. Data Overview:

• Q2 revenue of $96.2bn (+106% YoY, +18% QoQ), beating consensus estimates (~$92.2bn).
• Data center revenue of $89.0bn (+117% YoY, +18% QoQ).
• Adj. EPS of $2.22 (consensus ~$2.10).
• Gross margin remains at 75%, with adjusted operating margin at ~66%.
• Q3 guidance of ~$108bn (±2%), exceeding consensus of $104.2bn, assuming zero data center compute revenue from China.

Goldman Sachs quickly revised its earnings forecast post-earnings. They project NVIDIA's 2028 revenue will approach $955bn, up from the current median estimate of $790bn, a 20.9% increase.

In August, alongside storage companies enhancing shareholder returns, the combination of NVIDIA + CSPs/software + optical interconnects has remained a hot topic. As data centers continue to proliferate, demand for rack interconnects is amplifying, accelerating the ramp-up of NVIDIA's CPO switches.

On rack delivery, the new Rubin cards arrived early, ensuring no gap in the generational transition; Jensen Huang noted on the earnings call that while orders are plentiful, even a fully maximized supply chain can only meet 70% of demand.

2. The Overlooked AICE Segment

AICE Business (AI Clouds, Enterprise, Sovereign)

Jensen Huang mentioned that besides hyperscale cloud giants (Microsoft, Amazon, etc.), the AICE segment—which includes autonomous AI, new clouds, AI startups, and enterprise applications—accounts for roughly half of total business and is growing at 100% annually, with exploding demand for both closed-source and open-source solutions.

The value of NVIDIA's full-stack AI factory solution lies in serving these SME clients who lack the capability or willingness to self-develop chips like Google's TPUs, thus needing an 'out-of-the-box' AI platform.

New Cloud ETF:$Roundhill Neocloud ETF(NCLD.US)

Compute Infrastructure:$Nebius(NBIS.US)$Coreweave(CRWV.US)$IREN(IREN.US)$Hut 8 Mining(HUT.US)$Oracle(ORCL.US)

AI Server Deployment:$Super Micro Computer(SMCI.US)$Dell Tech(DELL.US)$Hewlett Packard Enterprise(HPE.US)

Leveraged ETFs:$Leverage Shares 2X Long NBIS Daily ETF(NBIG.US)$Leverage Shares 2X Long CRWV Daily ETF(CRWG.US)$IREN 2X Long ETF(IRE.US)$SMCI 2X Long ETF(SMCL.US)$DELL 2X Long ETF(DLLL.US)$ORCL 2X Long ETF(ORCX.US)

3. NVIDIA Unveils Custom Memory 'NVHBM': Bandwidth Up 30%, Power Down 15%

HBM consists of stacked memory layers that need to communicate with compute chips via a memory controller acting as a translator. Previously, this 'translator' sat on the compute chip, consuming valuable silicon area meant for computation.

In short: NVIDIA moved the memory 'translator' from its own GPUs to the base of the memory modules, turning the blueprint into a licensed revenue stream.

Post-move, NVIDIA reported three key metrics: 30% higher bandwidth, 15% lower HBM power consumption, and 25% more compute chip area freed up for processing.

The first major adopter is $Amazon(AMZN.US), which announced plans to deploy an additional 2 million NVIDIA GPUs by 2027–2028. NVIDIA expects AI Labs to contribute roughly a quarter of the company's overall business next year.

Self-developed chips vs. buying from NVIDIA isn't a 'pick one' choice; it's division of labor. Handle cost-saving tasks internally, but outsource complex needs to NVIDIA.

NVLink Fusion solves whether your custom chip fits into my rack; NVHBM solves whether your memory uses my IP.

NVIDIA now offers a 'standard implementation solution' that anyone can follow. The remaining competitive edge for the three players lies in DRAM process technology, stacking techniques, yield rates, and thermal management—these are hard skills ($SK Hynix(SKHY.US)$Micron Tech(MU.US) and Samsung).

Storage Sector ETFs:$Roundhill Memory ETF(DRAM.US)$Tema Memory ETF(DISK.US)$Defiance Daily Target 2X Long DRAM ETF(DRAL.US)

Other Storage Plays:$Seagate Tech(STX.US)$Sandisk(SNDK.US)$Western Digital(WDC.US)$CXMT(688825.SH)

Leveraged ETFs:$MU 2X Long ETF(MUU.US)$SKHY 2X Long ETF(SKHX.US)$SNDK 2X Long ETF(SNXX.US)$STX 2X Long ETF(STXL.US)$WDC 2X Long ETF(WDCX.US)

$XL2CSOPHYNIX(07709.HK)$XL2CSOPSMSN(07747.HK)