Investors should prioritize Anthropic as a top-tier AI play, focusing on its dominance in the $500 billion AI-assisted coding market and its transition toward autonomous "agentic" software. NVIDIA (NVDA) remains a high-conviction core holding as it shifts to 1-year innovation cycles, maintaining a structural lead amidst a multi-year global compute shortage. To capture the "unsexy" infrastructure boom, look to Celestica (CLS) for its Google TPU partnership and Corning (GLW) for the massive fiber optic demand required by new data centers. Alphabet (GOOGL) offers a safer entry into the AI arms race due to its vertical integration with internal TPU chips and superior data-handling capabilities via Gemini. Conversely, reduce exposure to traditional SaaS and "per-seat" software providers like Salesforce (CRM), as enterprise budgets are being diverted away from legacy applications toward AI tokens and hardware.

Anthropic

The speaker identifies Anthropic as their highest conviction investment, positioning it as a "dark horse" that has successfully carved out a dominant niche in the enterprise AI sector.

• Enterprise Focus: Unlike OpenAI, which won the consumer market early, Anthropic focused purely on enterprise applications.
• The "Coding" Unlock: The primary catalyst for conviction is Anthropic’s dominance in AI-assisted coding.
  • The market for AI coding tools is estimated at roughly $500 billion, based on 20 million global coders spending ~$20,000/year on tokens.
  • Anthropic’s model (Claude) has transitioned from a "copilot" (writing snippets) to being "agentic" (writing entire programs from English prompts).
• Ecosystem Strategy: Anthropic is moving beyond a simple API to build a "harness" of software, including SDKs and orchestration layers, creating a "lock-in" effect similar to AWS in its early days.
• Valuation Growth: The speaker noted the company's rapid revenue scaling, moving from $100 million to $1 billion, and potentially toward $9 billion in a very short window.

Takeaways

• Look for "Agentic" Capabilities: The next phase of AI value is not just answering questions, but performing autonomous tasks (agents).
• Coding as a Proxy: Use a company's performance in coding benchmarks as a leading indicator of its model's reasoning capabilities and enterprise utility.
• The Three-Horse Race: The foundational model layer is becoming an oligopoly consisting of OpenAI, Anthropic, and Google (Gemini).

NVIDIA (NVDA)

The speaker views NVIDIA not just as a chip company, but as the primary beneficiary of a "de-commoditization" of hardware.

• Exponential Earnings: The speaker highlighted that they were buying NVIDIA in 2023 at only 4x earnings because the market failed to model exponential growth correctly.
• Compute Shortage: There is a structural shortage of compute power. Venture capitalist Marc Andreessen is cited as saying there will not be enough compute for at least the next four years.
• Innovation Cycles: Unlike the old 7-year hardware cycles, NVIDIA is pushing the industry into 1-year upgrade cycles, forcing constant reinvestment from big tech.

Takeaways

• Don't Fear the "Chart": High-performing stocks like NVIDIA often look like "bubbles" on a chart, but if the earnings growth is exponential, the Price-to-Earnings (P/E) ratio can actually remain low.
• Hardware is the New Software: In the AI era, hardware has high intellectual property (IP) and "moats," making it a high-margin business rather than a commodity.

The "Infrastructure Stack" (Supply Chain)

Beyond the famous chip makers, the speaker identifies massive opportunities in the "unsexy" parts of the data center that are being pushed to their physical limits.

• Celestica (CLS): A contract manufacturer that transitioned from commodity electronics to specialized AI servers. They are a sole supplier for Google’s TPU and have a 50-60% share in the cloud Ethernet switch market.
• Corning (GLW): Benefiting from the massive need for fiber optics. One Microsoft data center requires enough fiber to circle the earth 4.5 times.
• Elite Materials: Provides copper-clad laminates for 40-layer printed circuit boards (PCBs) required for AI servers (standard servers only need 10 layers).
• Vertiv / Delta / Advanced Energy: Companies providing power supplies and liquid cooling. AI racks use 50% to 125% more power than traditional racks, driving up Average Selling Prices (ASPs).

Takeaways

• The "L-Curve" of Demand: While most technologies follow an S-curve (slow, then fast, then flat), the speaker describes AI infrastructure demand as an "L-curve"—going straight up with no sign of slowing.
• Identify Bottlenecks: Investment opportunities exist wherever there is a shortage (e.g., high-bandwidth memory, specialized PCBs, and liquid cooling systems).

Alphabet / Google (GOOGL)

Despite market concerns about Google being "behind" in AI, the speaker maintains it as one of their largest positions.

• Gemini’s Strength: Google is noted for having superior capabilities in "ingesting PDFs" and handling massive amounts of data.
• Infrastructure Advantage: Google’s internal TPU (Tensor Processing Unit) chips give them a vertical integration advantage that other software companies lack.

Takeaways

• Don't Count Out Incumbents: Companies with massive "cash cows" and existing distribution (billions of users) have the capital to survive the expensive AI "arms race."

Enterprise Software (SaaS) - Bearish Sentiment

The speaker has sold almost all positions in traditional application software (SaaS) companies, citing significant disruption risks.

• Budget Cannibalization: CIOs are diverting budgets away from traditional software (like Salesforce) to buy AI tokens and compute.
• Pricing Pressure: Traditional SaaS companies used to raise prices annually; AI makes it harder to justify these increases if the AI isn't significantly better than free or cheap models.
• The "Seat" Risk: If AI reduces the number of employees needed (e.g., in customer service or coding), software companies that charge "per seat" will see revenue decline.

Takeaways

• Wait for "AI Native" Winners: The speaker suggests it is currently safer to invest in the "shovels" (chips/power) than the "apps," as it is unclear which software companies will survive the transition.
• The "Rule of 40" for AI: Evaluate companies based on: (Percentage of Sales from AI) + (Market Share in that AI category). If this sum is high, the company is a strong candidate.

Investment Themes: The S-Curve Framework

The core philosophy discussed is identifying where a technology sits on the S-Curve of Adoption.

• Current Status: Enterprise AI is currently less than 1% penetrated.
• The "Tornado" Phase: We are entering the "Tornado" phase where barriers to adoption (price, complexity, security) are removed, leading to vertical growth.
• Alpha in Mega-Caps: The speaker argues there is significant "Alpha" (market-beating returns) in the largest tech companies because many generalist investors underweight them, failing to realize that in the digital age, the "winner takes most."