🚨10X Stocks - My New Book Published worldwide on April 6th Happy to share it’s already #1 New release and #9 Best seller in Business/Valuation on Amazon 😊 The book as the title says is about: 8 Frameworks on ‘How to Pick Multibaggers’ Order 👉 10X Stocks: How to Pick Multibaggers a.co/d/059DfXkj

📈OPTICS : JUS GETTIN STARTED (SECULAR)

The only trigger for $NKE would be China. The revenues has dipped by 10% over last 6 years!! Well $NKE trading at 4% FCF Yield and a 20% Sales growth at this low base can bring op-leverage to action. It’s cheap ‘now’

🚨 One of the cleanest high-frequency AI indicators I use to track the AI build-out is South Korea’s semiconductor export data 🇰🇷 It updates every 10 days, which gives you a read on the hardware cycle months before earnings. The latest print (Sep 1–20): 🔥 Chip exports: $34.1B, +259% YoY 📈 Record for any 20-day period (prior high: $25.5B in June) 🧩 Chips are now ~48% of ALL Korean exports 🚢 Total exports: $71.4B, +78% YoY Look at the trend in 2026 first-20-day chip exports: Apr +182% → May +202% → Jun +188% → Jul +180% → Aug +199% → Sep +259% This isn’t slowing down. It’s accelerating. The memory super-cycle can go much longer than we can remain skeptical : $MU $SNDK $DRAM 📈

There we go $PYPL $META potential

$NBIS: Why Nebius's 100% of contracted backlog will translate into revenue — and what Meta's share really looks like ?? 📰2 contracts form the backbone of Nebius's order book - $MSFT (25%) and $META (21%). US Neolab, Cohere and Reflection adds to remaining contracts. Contract-by-Contract Breakdown: Customer TCV Duration MSFT $19.4 B 5 Y (2031) META $12 B 5 Y (Early 2027) (dedicated) META (backstop) $15 B 5Y (rolling ) The consensus sees $12B in 2027 and $25B in 2028 for $NBIS Meta's contribution is estimated at ~15% of revenue in 2027 and much higher (25%) in 2028. Why? 'Backstop' Agreement with $NBIS. RPO expected to convert within 2 years: $13.5 B Contracted power (target 2026): 5 GW Key takeaway: The most credible range for 2027 revenue is $11-13B and for 2028 it is $19-27B (consensus ~$24.9B) And remaining $15 B RPOs flowing into 2028 and onwards. So $NBIS is now trading at 4X 2027 sales and 2X 2028 sales (90% conversion probability). Given the success virality of $META muse and Copilot 'super app' launch by $MSFT to monetize AI are strong long-term tailwinds to lock-up compute demand with $NBIS to meet scale 📈

$ORCL Oracle just sent a “force majeure” notice related to ‘Project Jupiter’ Reason: New Mexico denied Gas pipeline energy permit twice Oracle had contracted with Bloom Energy to deliver up to 2.45 gigawatts of fuel cells to power the software giant’s New Mexico data center project $BE Irony — fuel cells are sold as the way around grid bottlenecks, but they still need gas infrastructure. Here the constraint is a pipeline permit, not the grid !! ⛔️ Valuation makes it more sensitive. $BE is up about 230% over one year and trades at 22Xx sales, so any slip in the timeline gets punished.

🚨 Brilliant breakdown by @OrnnExchange on what ($NVDA) H100 and B300 owners actually earn (rent + resale, 80% util) 👇 H100 bought Sep 2025: 1 year in: +75.6% Projected by Sep 2028: ~$63,500 (+227%) B300: ~$186,000 (+244%) by Sep 2028 And the H100 still resells above what it cost a year ago. Hopper holds value. Now $NBIS is raising on-demand prices 17–21% from Oct 1 and 70% of its Q2 deals came with customer prepayments !! 📈

⛔️ $META MUSE JUST WENT VIRAL 🚀 40 Stocks Powering Meta’s MUSE & the Agentic AI Boom 🤖 AI agents take actions and that takes a ‘LOT ‘ more compute/power 📈 Here’s the full stack: LLMs / Agent Brains: $META $MSFT $GOOGL $AMZN CPUs (Agent Command Center): $INTC $AMD $ARM $QCOM GPUs / ASICs: $NVDA $AVGO $MRVL $TSM Memory: $MU $SNDK $DRAM Servers: $DELL $SMCI Optical Transceivers: $LITE $LAZR Fiber & Connectors: $APH Connectivity Chips: $CRDO $ALAB $MTSI Networking & Switches: $ANET $CSCO DCI: $CIEN $NOK Power & Cooling: $VRT $GEV $VST EDGe Compute & Security: $NET $S Agent Security: $CRWD $PANW $NET

🏮Optics is the future of AI Data Connect ! 📈Here are growth triggers for all 20 Optics stocks across segments— 1. Transceivers & Optical Modules:: $LITE Lumentum: AI datacom lasers are sold out. Management flagged an EML laser shortfall of roughly 25–30%, plans about 40% more unit capacity, and has optical circuit switch (OCS) and co-packaged optics (CPO) ramps coming as catalysts 🚀 It also landed the largest single purchase commitment in its history for ultra high-power lasers tied to CPO $COHR Coherent: This is a scale play on the 800G to 1.6T transceiver upgrade. Coherent makes its own indium phosphide (InP) lasers, which protects margins while peers are supply-constrained. $AAOI Applied Opto: posted record Q2 revenue of $191 M up 86% year on year, with 800G revenue more than doubling sequentially 📈 Management expects demand to outpace capacity through mid-2027 and targets about 650K units a month of 800G/1.6T by year-end. $FN Fabrinet: picks n shovels contract manufacturer behind the big optics and GPU-networking names. 2. Lasers & Optical Networking:: $IPGP IPG Photonics: This is more of an industrial and cyclical play than a pure AI one. The triggers are a manufacturing recovery, plus newer bets in defense, medical and micromachining. It is a top-5 holding in the photonics ETFs, so it benefits from theme flows. $LASR nLight: The main driver is defense directed-energy lasers, as laser weapon programs move from prototype to procurement. Its high-power semiconductor lasers also feed industrial and aerospace demand. $CIEN Ciena: As AI clusters span multiple data centers, the links between sites need coherent optics. That makes Ciena a “scale-across” winner, with its latest-generation WaveLogic modems and hyperscaler data-center interconnect spending as the triggers $VIAV Viavi: Every 1.6T and CPO product has to be tested. Viavi sells test gear into the speed upgrade cycle 3. Foundries & Materials:: $TSEM Tower Semi: This is a leading silicon photonics foundry for 800G/1.6T chips. Capacity expansions and new wins with transceiver makers are the triggers. $AXTI AXT: It supplies the InP substrates that every datacom laser is built on, so laser shortages translate directly into substrate demand. $AIXXF Aixtron: It makes MOCVD tools, the machines that grow the crystal layers for InP and GaAs lasers. Laser capacity build-outs across the industry feed its order book, along with GaN power chips. $SLOIF Soitec: Its silicon-on-insulator (SOI) wafers used for silicon photonics are the AI angle. Weakness in its larger mobile and auto markets has weighed on results, so this is a turnaround + photonics story. Emerging Photonics:: $POET POET Tech: Lumilens placed an initial $50M order for POET’s optical engines, under a framework that could reach more than $500M over 5 years. It holds over $830M in cash with little debt. Samples of its 1.6T module with Lessengers are targeted for Q3 2026. $LAZR Tema: It holds Anthropic through an SPV at about 12% of assets and was launched with SemiAnalysis. That makes it a photonics fund with pre-IPO AI exposure built in.

🚨 $META CONNECT : Mark doing a live demo of voice based interaction with Muse. It’s truly a game changer !! $Meta will need keep up or increase compute CAPEX to meet the adoption curve of Muse. Bullish for all AI ecosystem Agentic: $META Payments: $PYPL Stripe $CRCL Apps: $SHOP Neocloud : $NBIS $CRWV Memory: $DRAM Connects : $CRDO $ALAB $LITE 📈

Price Multiples doesn’t matter if you find a great earnings compounder that can grow profits at 25% CAGR for 20+ years !! In 1948 Graham’s investment partnership purchased 50% of GEICO for $712,000. Graham’s one-half purchased interest amounted to a purchase of 1,500 shares at $475 per share (a 10% discount to book value). Graham made a 500X.. story 👇

$GOOGL IS SO CHEAP — TPUs ORDER BOOK IS A LARGE NEW TAM 📈

🚨 There is a 80% high probability that $META could buy $PYPL and augment it to META-PAY It’s a direct call to bring 440 million global accounts into meta muse platform 📈

🚨 Anthropic launches Claude Opus 5.5 — now at Stanford level in Perplexity computer WANDR benchmark. It performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5 !! Claude Opus at $4.13 / task GPT Astra at $11.98 / task Claude Opus 5.5 is now available on $AMZN AWS , $SNOW Cortex

🚀 So glad $META integrated Muse with $PYPL @alexandr_wang 👍 Had tweeted this couple days back on why $PYPL is critical payment rail for Muse 👇

🚨MAG7 : All of them became Trillion Dollar Companies due to ‘Inflection points’. They could build ‘Optionality’ due to Asset-light /High Cash flow business models . Let’s deep dive to identify future MAG7 stocks👇 $GOOGL : Google’s Optionality I election point— TPU commercialization came in Oct 2025 with Anthropic’s deal for up to 1M TPUs. Today, Google Cloud revenue jumped 82% to $24.8B and TPU system sales are now reflected in Google Cloud’s $514B cloud backlog 📈 $NVDA: Once just a gaming/bitcoin GPU Seller, Nvidia’s inflection point was in May 2023, when it guided revenue 50% above Wall Street’s estimates as ChatGPT sparked the AI arms race. Datacenter GPUs became the optionality for $NVDA and many others like $AMD Today, NVDA Data Center revenue is $89 billion a quarter, up 117% year over year 📈 $MSFT: Microsoft’s 1st inflection point was Satya Nadella becoming CEO in February 2014 and pivoting Microsoft to “mobile-first, cloud-first (Azure).” Microsoft’s 2nd inflection point was in January 2023, when it made its multibillion-dollar bet on OpenAI. Today, Azure is a $100 billion+ business growing 43%, with a $678 billion commercial backlog 📈 $AMZN: Amazon’s latest inflection point was its ads business strategy in 2021, when it first broke out advertising as a $31 billion business. Today, Amazon Ads contributes a 50% of Amazon’s operating profit !! Ads is a $76 billion business for Amazon, with Q2 revenue up 26% to $19.8 billion The inflection was its AI Chips winning the top AI labs line Anthropic and OpenAI. Amazon’s chips business exceeded a $25B annual run rate, AWS growth accelerated for the fifth straight quarter to 36.7%, and the backlog stands at $496B $META: Meta’s next inflection point just happened in April 2026, when Meta Superintelligence Labs launched Muse Spark, its first model. Today, Muse is Meta’s first paid API for its own frontier model. MUSE just knocked ChatGPT off with 1.8M iOS downloads (US+CA) in 12 days vs ChatGPT’s 1.3M 📈 $TSLA : Tesla’s inflection point could be FSD when it moved to subscription-only and started Robotaxi rides Today, Tesla has 1.48 million active FSD users globally and Robotaxis have driven over 1 million unsupervised miles 📈

$META Muse just knocked ChatGPT off #1 on the US App Store 🤯 Launch vs launch: 📲 1.8M iOS downloads (US+CA) in 12 days vs ChatGPT’s 1.3M 👥 642K daily users vs ChatGPT’s 231K 🌍 2.8M installs total 4B users to funnel in. Distribution is the moat For $META Where the numbers come from: ✅ Downloads: Apptopia estimates 1.8M Muse iOS downloads in the U.S. and Canada in its first 12 days, versus 1.3M for ChatGPT over the equivalent post-launch period ✅ Daily users: 642K daily active mobile users in the U.S., compared with 231K for ChatGPT at the same point after launch. ✅ Total installs: Across all platforms and markets, Muse reached an estimated 2.8M installs in those first 12 days. ✅#1 ranking: Muse reached No. 1 on Apple’s US free iPhone chart on 18 September, ten days after launch ✅ Distribution: Meta has 4 billion users across Facebook, Instagram, WhatsApp, and Threads.

🚨 $NSCL IS THE DIRECT BET ON ANTHROPIC GROWTH !! Anthropic is Nscale’s largest customer - both IPOs to hit markets soon 📈 $NSCL Vs $NBIS (H1 2026): NSCALE : Revenue $140 M Op.Margin: (-) Profits: - $1 Billion NEBIUS : Revenue $980 M Op.Margin: +37% Profits: $430.8 M Why $NSCL is losing money ? 1.Costs >> revenue. Cost of goods sold excluding D&A was $189.6M in H1( above revenue). Sites, power and staff are paid for before the GPUs are billing 2.Currency, not operations. $457.1M, about 45% of the loss, came from foreign-exchange fair value losses. That’s non-cash !! 3.Customer concentration. One customer provided 52% of H1 revenue, so Nscale has little pricing leverage yet. 4. Active fleet: $2.6B (2.5%). Only this slice is generating revenue today. Nscale runs 25,000 active GPUs against 461,000 active or contracted, so about 5% of the fleet is live. $NBIS has - scale - diverse streams - Neocloud value add Softwares and - better pricing power

🚨 Alibaba Qwen is so underpriced . The best optionality for ‘open weight’ model is $BABA and Kimi ( Alibaba owns 35% ) Expect a MUSE like release from $BABA Soon !!

🚨 In 2025, inference was 9% of global data center workloads vs 14% for training. By 2030: inference 37%, training just 13%. AI is moving from building models to running them Here’s what that means for every chip giant and challenger in my Semiconductor map 🧵 1/ Compute $NVDA Still the default, but inference is where competition is fiercest. Cost per token matters more than raw speed $AMD Memory-heavy MI chips fit inference well. Biggest share-gain setup. $INTC , $ARM Agentic AI needs far more CPUs. MediaTek: edge + custom ASIC work. 2/ Custom silicon & networking : $AVGO Inference is where custom ASICs beat GPUs on cost, so every hyperscaler wants its own chip $MRVL : CXL, ASICs and Optical scaling for data inference $ANET Inference spreads across many clusters, which means more front-end networking $CRDO More racks mean more AI inter-connections. AECs and optics scale with every deployment. 3/ Memory, the quiet winner: Inference is memory-bound: long context, KV cache, agents holding state. $MU , Samsung, CXMT: HBM + DRAM demand stays sanguine for inferencing. $SIMO KV cache is spilling onto SSDs, so enterprise SSD controllers matter. 4/ Foundry & equipment $TSM Wins either way: GPU, ASIC or CPU, it gets fabbed in Taiwan. $LRCX $KLAC More chips means more fabs. Memory capex favors ($LRCX) etch especially. 5/ Challengers $CBRS - A pure play on fast inference. This shift is its entire thesis. $BABA T-Head: Alibaba’s in-house chips serving Qwen at China scale. Watch for the IPO