Investors should prioritize NVIDIA (NVDA) as it expands beyond GPUs into the CPU market with its Grace chips, targeting $20 billion in new revenue to capture the shift toward "Agentic AI." For high-upside growth, AMD is a strong conviction play with a path to $20 earnings per share by 2028 as it aggressively gains server market share and secures major GPU deals with Meta and OpenAI. Broadcom (AVGO) offers a resilient infrastructure play, combining massive AI networking growth with steady cash flows from its recent VMware acquisition. To capitalize on the critical hardware bottlenecks, consider a "basket" approach to semiconductor equipment makers like ASML and Lam Research (LRCX), or high-performance memory leader Micron (MU). Be mindful of the primary sector risk: any significant reduction in AI capital expenditures from "hyperscalers" like Google or Microsoft would signal a time to exit these positions.

NVIDIA (NVDA)

• NVIDIA is described as the "center of the universe" for AI infrastructure, with revenue growth accelerating from 65% to 85% in recent quarters.
• The company has resegmented its business to show that data center revenue is split roughly 50/50 between "hyperscalers" (Google, Meta) and "non-hyperscalers" (enterprise, sovereign states, and "neoclouds").
• NVIDIA is aggressively entering the CPU market with its "Grace" chips, projecting $20 billion in CPU revenue this year—a scale comparable to Intel or AMD.
• Management has been highly effective at securing supply across the value chain (TSMC, memory, packaging) to meet accelerating demand.

Takeaways

• Valuation Opportunity: Despite massive growth, the stock has lagged some peers recently. Eisman and Rasgon suggest this is due to technical factors (institutional ownership limits) and investors "playing the bottlenecks" in other sectors.
• CPU Expansion: Investors should watch for the shift from "Generative AI" to "Agentic AI," which requires more CPU power. NVIDIA is positioning itself to capture this market, not just the GPU market.
• Bullish Outlook: The upcoming "Blackwell" and "Rubin" chip architectures are expected to maintain the company's lead.

Advanced Micro Devices (AMD)

• AMD has successfully taken massive market share from Intel in the x86 server CPU market, moving from 0.1% share 10 years ago to over 40% today.
• The company is using warrants to secure large GPU deals with OpenAI and Meta, essentially "buying their ticket" to build a developer ecosystem.
• CEO Lisa Sue recently doubled the 2030 addressable market estimate for AI accelerators from $60 billion to $120 billion.

Takeaways

• Earnings Potential: Rasgon suggests AMD could reach $20 in earnings per share by 2028 if current trends continue, which would support significant stock price upside.
• Strategic Trade-offs: While the use of warrants is dilutive, it is viewed as a necessary move to challenge NVIDIA's dominance in the GPU ecosystem.

Intel (INTC)

• Intel is currently benefiting from a "rising tide" where demand is so high that customers are buying even their less competitive, older server products.
• The company is pivoting toward a "foundry" model (manufacturing chips for others) and has improved its balance sheet by exiting a dilutive deal with Apollo.
• There are rumors of Intel securing foundry customers like Apple, and they possess valuable intellectual property in "packaging" (combining multiple chips into one unit).

Takeaways

• Execution Risk: While the narrative is improving under CEO Pat Gelsinger and the new leadership, the company still has "a lot of wood to chop" regarding manufacturing yields.
• Political Tailwinds: The transcript notes that political support for domestic chip manufacturing (specifically mentioning Trump's interest) provides a floor for the stock's sentiment.

Broadcom (AVGO)

• Broadcom is a major player in AI networking and "custom chips" (ASICs), working with Google on their TPU (Tensor Processing Unit) for over 15 years.
• The company recently acquired VMware, adding a massive software component to its business.
• Their AI-related business is projected to reach $100 billion next year, which is larger than the entire company was just a few years ago.

Takeaways

• Infrastructure Play: Unlike "SaaS" software companies that might be disrupted by AI, Broadcom’s software (virtualization) is the infrastructure that AI runs on, making it more resilient.
• Diversified Exposure: Broadcom offers a mix of semiconductor growth and steady software cash flows, though the stock has recently lagged due to broader weakness in the software sector.

Micron (MU) & Memory Sector

• Memory stocks like Micron have outperformed (up well over 100%) because memory has become a major bottleneck in AI server production.
• The industry is highly cyclical; currently, earnings are exploding because pricing for high-performance memory has "gone crazy."

Takeaways

• Cyclical Profits: Memory companies are currently trading at single-digit P/E multiples because investors know the industry is cyclical. However, as long as AI demand accelerates, these earnings may stay elevated longer than expected.

Semiconductor Equipment (AMAT, LRCX, KLAC, ASML)

• These companies make the machines that build the chips. The "Big Five" (Applied Materials, Lam Research, KLA, ASML, and Tokyo Electron) control over 70% of the market.
• ASML holds a near-monopoly on the most advanced lithography machines required for high-end AI chips.

Takeaways

• The "Basket" Approach: These stocks tend to be highly correlated. If the semiconductor sector is growing, the equipment makers almost always benefit. Lam Research (LRCX) has recently been a top performer in this group.

Investment Themes & Risks

The Rise of "Agentic AI"

• The next phase of AI involves "agents" that perform real-world tasks (e.g., booking travel, writing and testing code).
• Insight: This shift requires significantly more CPU power compared to simple text generation, benefiting NVIDIA, AMD, and Intel.

The "Return on Investment" (ROI) Bear Case

• The primary risk to the boom is if big tech companies (Google, Meta, Microsoft) do not see a financial return on their massive capital expenditures (CapEx).
• Risk Factor: If Meta or Google cuts their CapEx budget, the entire sector will likely experience a sharp downturn. However, current data shows AI revenue (e.g., Anthropic) is growing at an unprecedented pace.

The Power Constraint

• A physical limit to the AI boom is the electrical grid. The U.S. grid may not be able to support the projected data center growth.
• Opportunity: This is driving interest in on-site power generation, gas turbines, and Small Modular Reactors (SMRs). Companies like Microsoft are even looking at restarting nuclear plants (e.g., Three Mile Island).