Investors should consider AMD as a high-conviction play on the shift toward "Agentic AI," where the increasing demand for high-performance CPUs alongside GPUs is expected to drive long-term margin improvements. For exposure to the data infrastructure layer, MongoDB (MDB) is positioned as a critical winner as Fortune 100 companies transition from AI experimentation to production-grade applications. While currently private, keep a close watch on SpaceX for any secondary market opportunities or a future IPO, as analysts project it could eventually reach a $10 trillion valuation as the foundational "railroad" of the space economy. In the hardware sector, look for emerging opportunities in Edge AI and companies like SambaNova Systems that specialize in "Premium Inference" to reduce the high power costs associated with traditional GPUs. Finally, monitor the shift toward Open Source models, which are predicted to capture 90% of AI token volume within the next 12 months, favoring cost-efficient production workflows over expensive closed-source alternatives.

AMD (AMD)

• Mark Papermaster (CTO) highlighted the massive shift toward Agentic AI, which is transforming end-to-end workflows and chip design.
• Hardware Shift: In agentic workflows, the ratio of CPU to GPU is moving toward 1:1, requiring high performance from both components.
• Internal Efficiency: AMD is currently using thousands of "sub-agents" to handle complex chip design, significantly accelerating their time-to-market for next-generation hardware.

Takeaways

• Bullish on Diversified Compute: Investors should note that while GPUs are the stars of AI, the rise of autonomous agents is re-validating the importance of high-performance CPUs.
• Efficiency Gains: AMD’s internal use of AI for chip design suggests a long-term margin improvement and faster innovation cycles compared to traditional manual design processes.

SpaceX (Private)

• Philip Johnston (CEO of StarCloud) claims the SpaceX IPO (whenever it occurs) will historically be viewed as the "most undervalued IPO of all time."
• Market Dominance: SpaceX owns the most cost-effective launch vehicle, making them the "railroad" of the space economy.
• Total Addressable Market (TAM): Beyond launch, the company is positioned to capture value in asteroid mining, lunar resources, and orbital communications.

Takeaways

• Long-term Valuation: Johnston predicts the company could eventually surpass a $10 trillion valuation.
• Infrastructure Play: SpaceX is viewed as the foundational layer upon which all other space-based businesses (like orbital data centers) must be built.

SambaNova Systems (Private)

• Series F Funding: The company recently announced a $1 billion raise at an $11 billion valuation, led by General Atlantic.
• SM50 Chip: Launching later this year, this "RGU" (Reconfigurable Dataflow Unit) is designed for "Premium Inference."
• Competitive Edge: Claims to deliver higher performance for trillion-parameter models at a fraction of the cost and power consumption of traditional GPUs.

Takeaways

• Inference Era: As the industry shifts from training models to running them (inference), SambaNova is positioning itself as a high-margin alternative to NVIDIA for enterprise-scale applications.

MongoDB (MDB)

• AI Integration: Mentioned as a critical layer for storing and reasoning over data for AI agents in real-time.
• Market Position: Currently used by 75% of the Fortune 100; the platform integrates vector search and embeddings (via Voyage AI) to eliminate the need for complex, multi-tool data pipelines.

Takeaways

• Infrastructure Winner: As more companies move from AI experimentation to production-grade "agentic" apps, specialized databases like MongoDB are expected to see significant growth.

Investment Themes & Sectors

1. The Death of the "GUI" (Keyboard and Mouse)

• Max Cook (Coatue) suggests we are moving from a world of 60-80 apps to an "Agentic AI" world where users interact with only 2-3 "super apps" via natural language.
• Investment Insight: Bullish on wearables and natural language processing (NLP). The focus is shifting from graphical user interfaces (GUIs) to "Thinking Machines" interaction models with zero latency.

2. Open Source vs. Closed Source Models

• The Shift: Matan (Factory) predicts that 90% of AI tokens will go to open-source models within 12 months.
• Production vs. Discovery: Open source (like Llama or Mistral) is becoming the standard for "production-scale" workflows due to low latency and cost, while closed frontier models (OpenAI, Anthropic) are used for "discovery" and the hardest tasks.
• Model Routing: A new sector is emerging for "model routers" that dynamically switch between cheap open-source models and expensive frontier models to optimize costs.

3. Physical AI and Edge Computing

• Liquid AI: Focusing on "Liquid Foundation Models" that are small and efficient enough to run on Raspberry Pis, cars (Mercedes-Benz), and airplanes rather than massive data centers.
• Investment Insight: Look for opportunities in Edge AI where intelligence is hosted on the device (substrate) rather than the cloud to save on energy and latency.

4. The "Clipping" Economy

• Good Future Media: Highlights a shift in digital marketing where "clipping" (short-form video hooks) is replacing traditional ads.
• Investment Insight: High demand for human-AI hybrid creative agencies that understand the psychology of the "3-second hook" to drive traffic to long-form content.

Risk Factors Mentioned

• Energy and Power Constraints: Ramin (Liquid AI) noted that energy is not abundant and the "maximize intelligence at all costs" strategy is hitting a wall due to grid bottlenecks.
• Supply Chain Disruptions: Specific mentions of memory supply chain issues and government control over AI assets as potential hurdles for foundation model companies.
• Open Source Decline in Spend: Despite high usage, enterprise spend on open source actually dropped from 19% to 11% recently, suggesting that while tokens are moving to open source, the revenue is still heavily concentrated in closed providers—for now.