OptionsPlay

Tony discusses the concerns over rising inflation after the latest PPI, before sharing a bullish opportunity in media and a bearish trade in the airline industry on this week’s #InTheMoney: bit.ly/3OlpLZR. Sponsored by @Fidelity

Most AI research tells you to buy NVIDIA and calls it a day. We scored 8 capabilities across every major tech company. NVIDIA got 2 out of 8. Great at chips, but no cloud, no consumer products, no frontier models. The company that scored 8/8 self-funds its entire AI buildout. $73B in free cash flow after $93B in capex. And the market thinks it's losing. We don't just rate stocks. We publish the methodology. Challenge any score, we'll show you the math. See the full scorecard → optionsplay.com/ai-page

When a sell-side analyst says "buy NVIDIA," they give you a price target and a thesis. But what they don't show you is where NVIDIA sits in the $675 billion AI infrastructure buildout. Who depends on NVIDIA, who does NVIDIA depend on, and which companies benefit even if NVIDIA loses market share. Every OptionsPlay research report starts with an ecosystem map. We show you the full value chain, how every company connects, where the bottlenecks are, and where value accrues. Then we score every company using 5 weighted factors calibrated for the specific opportunity. Our ratings adapt to what's actually happening in markets. Understanding how an industry works is worth more than any single stock pick. That understanding is what we deliver. Check out our AI Semiconductor research at the link below. optionsplay.com/ai-page

Everyone talks about NVIDIA when they talk about AI chips. But NVIDIA doesn't make a single chip. Every GPU they design has to be manufactured by one company: TSMC. The biggest bottleneck in AI right now isn't chip design. It's advanced packaging. TSMC's CoWoS technology is the only way to connect GPU chiplets with high-bandwidth memory. Capacity is ramping from 35,000 to 130,000 wafers per month, but NVIDIA alone is taking over 60% of it. That bottleneck affects every company in AI. NVIDIA, AMD, Broadcom, Google. All of them are constrained by the same packaging line. In our Semiconductor Deep Dive, we rated 25 companies across the entire chip supply chain. The names most people have never heard of might be the most interesting. See the full research at the link below. optionsplay.com/ai-page

Here's what inference actually costs per agent task: DeepSeek V3: $0.05 Claude Sonnet: $0.81 GPT-5.2: $0.63 Sounds cheap. Now multiply it by billions of daily tasks across every enterprise on the planet. 75% of knowledge workers use AI weekly. 84% of developers use AI coding tools. 80% of Fortune 500 companies have active AI agents deployed. Every single one of those interactions is an inference call. Every inference call requires compute, memory, power, and cooling. The shift from training to inference isn't just a technical detail. It's the demand driver that makes the entire AI infrastructure buildout necessary. Full breakdown in our research → optionsplay.com/ai-page

Here's the stat that should keep every SaaS CEO up at night: 75% of knowledge workers now use AI weekly, but 97% of them pay less than $20 a month. Goldman Sachs looked at this gap and said tech companies are overspending on AI. But they're measuring the wrong thing. An AI agent can now do the work of a $300,000/year analyst for about $2 to $3 per task. When enterprises figure that out, they won't pay $20 a month. They'll pay thousands. And the infrastructure demand to serve that usage will be unlike anything we've ever seen. The companies building that infrastructure are the ones we're buying. Full research available at the link below. optionsplay.com/ai-page

We scored every major tech company across 8 dimensions of the AI value chain. Custom silicon. Cloud infrastructure. Frontier models. Consumer distribution. Enterprise platform. Research. Cash flow self-funding. Data advantage. One company scored 8 out of 8. The only company competing at the highest level across every single dimension. NVIDIA: 2/8 Microsoft: 5/8 Meta: 5/8 OpenAI: 5/8 Oracle: 3/8 The consensus says this company is losing the AI race. We think the selloff just created the most compelling entry point in years. See which company scored 8/8 → optionsplay.com/ai-page

Meta just signed a deal with AMD worth up to $100 billion over 5 years. That's the largest semiconductor partnership in history. Why? Because Meta doesn't want to be 100% dependent on NVIDIA for inference compute. The Helios chip is custom designed specifically for running AI models at Meta's scale. AMD trades at just 31x forward earnings with 49% EPS growth and 35% revenue growth. That's cheaper than the semiconductor industry average. We rated AMD a Strong Buy in our Semiconductor Deep Dive. 25 companies. 9 sub-sectors. Every layer of the AI chip supply chain analyzed. Want the full analysis with every semiconductor stock ranked by AI exposure? Link below. optionsplay.com/ai-page

The biggest shift happening in AI right now isn't about training bigger models. It's about running them. 2024: Training was 65% of AI compute. Inference was 35%. 2026: Inference is now 67%. Training dropped to 33%. That inversion changes everything. Training a model is a one-time cost. Running it is continuous, compounding demand that scales with every new user and every new agent. DeepSeek, Claude, GPT. Every single API call generates real infrastructure revenue. As agents scale from millions to billions of daily tasks, inference revenue scales with it. This is why $675B in capex still isn't enough. Get access to the full research → optionsplay.com/ai-page

When you ask ChatGPT a question, that uses about 500 tokens of compute. One quick round trip. But when an AI coding agent fixes a bug, it reads the entire codebase, plans the fix, writes code, tests it, rewrites, and tests again until it works. That single task uses up to 1.5 million tokens. That's 3,000x more compute. Same user, same provider, completely different infrastructure demand. These products aren't prototypes. Claude Code hit $1B in revenue in 6 months. Cursor went from $1M to $1.2B in under 2 years. Wall Street is still projecting AI compute based on chatbots while the industry is building autonomous agents. That gap is your opportunity. We mapped 8 industries and 60+ companies positioned to capture this. Get access to the full research below. optionsplay.com/ai-page

Week ahead: This week we’re breaking down the most underappreciated bottleneck in the entire AI infrastructure buildout: power. Global data center electricity consumption: 448 TWh in 2025. Projected: 980 TWh by 2030. That’s a near doubling in 5 years. US data centers are on track to consume 7.8% of total electricity generation. Up from 3% just three years ago. New video + deep dive dropping this week. Don’t miss it. Subscribe → optionsplay.com/ai-page

Which layer of AI infrastructure has the most investment potential over the next 18 months? (We mapped all 8. The answer might surprise you.) Get the Full 8-Layer Breakdown ➡️ optionsplay.com/ai-page

On this week’s #InTheMoney, Tony discusses the continued Iran war and the impact that oil prices have on inflation, small-caps, and more. Watch now for his full analysis: bit.ly/3OlpLZR. Sponsored by @Fidelity

Everyone is asking: “Is AI a bubble?” Our answer: AI isn’t a bubble. It’s a bottleneck. The $2 trillion wiped from tech stocks isn’t a sign AI is failing. It’s capital transferring — from companies that resell AI features to companies that own the physical infrastructure AI depends on. Legacy SaaS at 10–15x revenue? Repriced to 3–5x. Infrastructure stocks? Record revenue, expanding margins. This rotation is structural, not cyclical. And it has a multi-year runway. Get the Full AI Infrastructure Thesis → optionsplay.com/ai-page

The math that breaks every AI model on Wall Street: Today: 75% of knowledge workers use AI as a chatbot. Each interaction = ~500 tokens. Tomorrow: Those same workers use AI agents. Each interaction = 75,000 to 1,500,000 tokens. Same users. 150x to 3,000x more compute demand. Without adding a single new user. Claude Code: $1B annual revenue in 6 months. Cursor: $1M → $1.2B revenue in under 2 years. 84% of developers now use or plan to use AI coding tools. This is not linear growth. It’s exponential. And the infrastructure isn’t ready. Get the Full AI Infrastructure Thesis → optionsplay.com/ai-page

Layer 1 of our AI Infrastructure thesis: Semiconductors & Compute. Every AI workload starts with silicon. Custom accelerators, GPUs, and application-specific chips are the picks and shovels of this cycle. Demand is outstripping supply across every major chip category. Lead times: 12–18 months for advanced AI processors. NVIDIA alone did $51.2 billion in data center revenue in a single quarter. $500B+ in outstanding orders for Blackwell and Rubin. This is Layer 1 of 8. Our research maps the full AI infrastructure stack. Unlock the Full 8-Layer Playbook → optionsplay.com/ai-page

Wall Street says hyperscalers are overspending on AI. Here’s what they’re missing: AI agents use 3,000x more compute than chatbots. That’s not a typo. A single autonomous coding session consumes 1.5 million tokens vs. 500 for a chat query. The market is modeling linear demand for a technology that scales exponentially. $675B+ in hyperscaler capex this year. $80B just six years ago. An 8.4x increase. And it’s still not enough. We mapped 8 industries, 30+ companies, and the most mispriced large-cap in the market. Get the Full AI Infrastructure Thesis → optionsplay.com/ai-page

Tony discusses whether markets have stabilized after a volatile week before sharing two tech trade ideas on this week’s #InTheMoney: bit.ly/3OlpLZR. Sponsored by @Fidelity

On this week’s #InTheMoney, Tony discusses the FOMC meeting, rising geopolitical tensions, and tech’s renewed role in the rally before sharing potential trade ideas in AI infrastructure: bit.ly/3OlpLZR. Sponsored by @Fidelity

With rates back in focus, we discuss why CPI could be the next big catalyst on this week’s #InTheMoney. Watch now for the full conversation: bit.ly/3OlpLZR. Sponsored by @Fidelity