Lian Loo @Chongloo: "🚀 NVIDIA GTC 2026 Key..."

Singapore

🚀 NVIDIA GTC 2026 Key Takeaways + How Big Is the Opportunity? + NVDA Trade Plan At the NVIDIA GTC 2026, Jensen Huang wasn’t just talking about better chips — he was redefining the entire business model of AI. His message was clear: 👉 AI is shifting from selling chips to building AI production systems. 🧠 1. AI Has Entered the “Inference Era” AI used to have two stages: Training Generation Now, we’ve officially entered the third stage: 👉 Inference This is a major shift because: Training = one-time Inference = happens every day Every AI query, coding task, automation, or enterprise workflow relies on inference. 📈 Implication: Compute demand → no longer cyclical Now → continuous, exponential, and monetizable Jensen projects that by 2027, AI infrastructure will be a $1T+ market. 🏭 2. The Rise of “Token Factories” A core concept introduced: “Future data centers are not computing centers — they are Token factories.” Tokens = the fundamental unit of AI Cost & revenue are measured in Tokens 📊 Future competition comes down to: Tokens per watt (efficiency) Token throughput (speed) 👉 AI becomes an industrial business 💰 3. AI Becomes a Commodity Market Pricing is already tiering: Low-end: a few dollars / million tokens High-end: $100+ / million tokens 📌 Key shift: 👉 AI is no longer just a cost center 👉 It becomes a priced, monetizable commodity Think: Electricity ⚡ Bandwidth 🌐 Oil 🛢️ ⚙️ 4. The Real Edge: System-Level Innovation Instead of relying on node shrink (e.g., 3nm → 2nm), NVIDIA is focusing on full-stack systems. Example: Vera Rubin system Same 1GW data center 👉 ~350x increase in Token output within 2 years 📉 Compared to: Moore’s Law ≈ 1.5x 👉 The winner is no longer the best chip 👉 It’s the best integrated system (GPU + networking + software + orchestration) 🔄 5. Business Model Shift: CAPEX → OPEX Before: Buy GPUs (one-time CAPEX) Now: Consume compute daily (OPEX) 📊 For NVIDIA: More recurring revenue Less cyclical Closer to a utility model 🤖 6. AaaS: Agent as a Service Software is evolving into AI employees: Customer service Reporting Decision-making Workflow automation 📌 Future companies may allocate: 👉 “Token budgets” like salaries AI compute becomes: 👉 A core operational necessity (like electricity) 🛰️ 7. The Next Frontier: Space Computing Thor chips are radiation-certified Already running in satellites Vera Rubin Space-1 in development Why space? ☀️ Near-unlimited solar energy 🌍 No land or grid constraints Even Elon Musk and SpaceX are aligned with this direction. 👉 Compute is becoming a strategic, energy-like resource 📊 8. Investment Framework: 3 Stages of AI 1️⃣ Sell the tools (chips) → $NVDA (NVIDIA Corporation) / $AMD (Advanced Micro Devices Inc) / $TSM (Taiwan Semiconductor Manufacturing Co Ltd - ADR) 2️⃣ Rent the tools (cloud) → $MSFT (Microsoft) / $AMZN (Amazon.com Inc) / $GOOGL (Alphabet Inc Class A) / $ORCL (Oracle Corporation) 3️⃣ Infrastructure (utility model) → Pay-per-use (Tokens) 👉 NVIDIA is expanding across all three layers. 💡 9. Valuation Re-rating Opportunity Current metrics: Forward P/E ≈ 22x EPS CAGR ≈ 41% PEG ≈ 0.53 📉 This reflects a hardware company valuation But if the market re-rates NVIDIA as: 👉 AI infrastructure / utility platform 📈 Then valuation could shift toward SaaS multiples (PEG ~1.5) ➡️ Multiple expansion = key upside driver 📌 NVDA Trade Plan Current levels to watch: Support 1: $170 Support 2: $150 👉 Strategy: scale in gradually 👉 Thesis: valuation compression + re-rating potential 🧾 Final Take 👉 AI is evolving from technology → industry → energy 👉 And NVIDIA is positioning itself as the power plant of the AI era $SPX500 $VOO (Vanguard S&P 500 ETF) $NSDQ100 $QQQ (Invesco QQQ)

Not investment advice. The author may have financial interests in the mentioned instruments.