Equity Empire Podcast

Everyone knows about the AI chip shortage. But after sitting through a dozen conference calls at the Morgan Stanley TMT conference, four semiconductor executives — from NXP, Analog Devices, ON Semi, and Texas Instruments — all pointed to the same trend nobody is talking about. TSMC is converting its older fabs to meet AI demand, and that's quietly draining supply from the analog chip market. These are the chips that go into cars, drones, robotics, wearables, and almost every electronic device. Inventories are at rock bottom, demand is about to recover, and most manufacturers won't be ready. One company spent six years and billions of dollars preparing for exactly this moment. We break down the setup, the risks, and why this could be the next big semiconductor trade.

Broadcom just dropped a bombshell on their latest earnings call — and Wall Street is still trying to figure out what hit them.The company quietly revealed they're tracking toward $100 billion in AI revenue. Not someday. Not in a decade. Within the next three years. And behind that number? Six customers. Just six. The most elite AI buyers on the planet, each building million-chip clusters that will reshape how we think about computing infrastructure.In this episode, we break down everything Hock Tan said on the call — and more importantly, what he didn't say. We go customer by customer through Broadcom's six hyperscaler relationships: Google, Anthropic, Meta, OpenAI, and the two mystery buyers Wall Street can't stop speculating about (ByteDance and Apple are the leading candidates). Each one tells a different story about where AI infrastructure spending is headed.We dig into why AI spending is irreversible once you've made the switch, why training isn't dead despite what the DeepSeek crowd keeps saying, and how the demand picture has actually gotten stronger — not weaker — since the last earnings cycle. Meta alone is scaling from hundreds of thousands of accelerators to millions. Anthropic is talking about 3 gigawatts of power demand. These aren't science projects. These are the biggest infrastructure buildouts since the internet itself.Then we do the math on Broadcom's path to a $3 trillion market cap, why Wall Street's current models are still underestimating the opportunity, and what this all means for investors watching the AI infrastructure trade play out in real time.Whether you're long Broadcom or just trying to understand the AI capex cycle, this episode connects the dots between earnings calls, datacenter buildouts, and where the smart money is actually flowing.0:00 Intro0:20 Broadcom's $100B Bombshell1:22 Why AI Spending Is Irreversible1:53 Hock Tan's Earnings Call4:18 Not a Zero-Sum Game5:57 Why Training Still Matters7:48 Training vs Inference: Two Tracks9:48 Demand Durability11:14 The 6 Elite Customers12:53 Customer #1: Google17:12 Customer #2: Anthropic20:31 Anthropic's 3GW Scaling21:31 Customer #3: Meta25:09 Meta's Two-Track Strategy26:47 Customer #4: OpenAI28:29 Customers #5 & #628:59 The Math Behind $100B31:56 Wall Street's Disconnect33:53 Key Investor Insights36:10 The Road to $3 Trillion38:00 Outro

Meta, Microsoft, Amazon, and Google are pouring over $1 trillion into AI infrastructure. I break down each company's strategy, rank them 1–4, and explain why one of them might never earn it back.0:00 Intro — The $1 Trillion AI Question0:29 Subscriber Jamie's Question2:22 GPU Real Estate Model — How Cloud CapEx Works4:26 META — All-In on AI (Not Renting GPUs)7:30 Meta's Hidden Opportunity Wall Street Is Missing8:25 Zuckerberg's Vision: Social Media Becomes a Life Tool13:34 Meta's Competitive Moat vs TikTok & Snapchat17:21 Meta's Power Move — Building the Next Era19:35 Why This Isn't the Metaverse Again23:11 MICROSOFT — The Azure & Copilot Play27:46 Microsoft's Familiar Playbook (Teams 2.0?)31:15 Can Copilot Become the AI Orchestration Layer?35:50 Enterprise Lock-In: Microsoft's Safety Net37:50 The Copilot Question — Will They Pull It Off?39:28 AMAZON — Instant Demand for Every GPU43:27 Amazon's Secret Weapon: They Don't Need Frontier AI45:13 The Golden Goose — $500B E-Commerce + AI48:39 GOOGLE — The Only Fully Vertically Integrated Stack52:07 Google's AI Specialization: Video, Images & DeepMind55:38 The Future of Ads — Why Clicks Get More Valuable58:25 RANKINGS — #1 Amazon, #2 Meta1:00:33 RANKINGS — #3 Google, #4 Microsoft1:03:47 Outro & How to Submit Questions

Wall Street is lumping all software stocks together and selling them off. That's a mistake. I walk through the 4-box checklist I use to find the software stocks that will actually recover — and which ones to avoid.

0:00 — Snowflake replaces Salesforce CRM, saves $5M/year
01:22 — Wall Street lumping all software into one basket
02:17 — Software stock charts breaking down (Salesforce, ServiceNow, Adobe, Trade Desk)
04:08 — The vibe coding bear argument
06:22 — Why Wall Street is making a big mistake
08:02 — Legacy software vs. the new infrastructure layer
10:16 — Software that actually does the work vs. UI software
15:08 — Snowflake proves growth without headcount (30% revenue growth, +37 employees)
17:39 — "Growth is no longer tied to headcount"
23:02 — The 4-Box Checklist begins
24:18 — Box 1: Where does the software live in the stack?
28:01 — Box 2: Is AI adoption accelerating?
31:21 — Box 3: Is revenue decoupling from headcount?
34:03 — Box 4: Does the chart confirm it?
40:54 — Recap and key takeaways

Wall Street is obsessed with legacy software and the cloud data center. But the agentic layer is guaranteed to replace humans inside organizations — and that work isn't going to the cloud. I break down why on-prem hardware is about to have its biggest moment in 15 years, which companies benefit (Apple, Nvidia, Dell, Intel, AMD, IBM, Oracle), and why the choice between renting the cloud and owning your own AI will be obvious.