Consider NVIDIA (NVDA) as a core, high-conviction foundational holding over the next three years, supported by massive hardware demand and major enterprise partnerships running through 2027–2028.

Buy Tesla (TSLA) for high-upside exposure to autonomous transport and robotics, keeping a close watch on the development of its in-house TerraFab chip facilities over the next two to three years.

Hold Palantir Technologies (PLTR) to capture enterprise software demand, as its stringent government security clearances create an exceptionally wide competitive moat.

Gain exposure to orbital computing and the commercial space theme via public launch provider Rocket Lab USA (RKLB) or private shares of SpaceX, which is co-designing space-grade compute hardware targeted for late 2027.

Diversify across the AI Infrastructure, Power, & Compute theme by investing in high-voltage electrical equipment and energy producers, which represent the most critical physical bottlenecks for long-term data center expansion.

NVIDIA (NVDA)

• Chosen by guest Hans as a top high-conviction, three-year holding among large-cap technology stocks.
• NVDA is seen as trading at a more attractive valuation relative to its long-term future than other high-profile tech peers.
  • The company recently reported 100% year-over-year growth and is guiding for approximately 70% year-over-year growth despite having a multi-trillion-dollar market capitalization.
  • CEO Jensen Huang is viewed as an elite operator actively resolving supply chain, memory, and physical infrastructure bottlenecks to maintain 50%+ growth.
• Tesla and SpaceX have reportedly secured 30% to 40% of the upcoming Vera Rubin chip supply for the next few years.
• NVIDIA is co-designing space-grade AI chips with SpaceX targeting deployment by late 2027.
• Key risks mentioned include execution risk at massive scale and high gross margins (around 85%) that incentivize customers to develop in-house custom silicon alternatives over time.

Takeaways

• Consider NVDA as a core foundational holding for exposure to the AI hardware build-out, with strong demand tailwinds through at least 2027–2028.
• Monitor the pace of customer custom-silicon initiatives (like Tesla's TerraFab) and memory supply constraints as long-term competitive risks.

Tesla (TSLA)

• Highlighted as having exposure to the largest total addressable markets (TAMs) in technology: autonomous driving (Cybercab/Robotaxi) and robotics (Optimus).
• Tesla is planning a dedicated chip manufacturing effort (TerraFab, starting with a mini-TerraFab test facility) to produce custom inference and memory chips at significantly lower costs than merchant silicon.
  • The goal is to design inference chips optimized for Optimus humanoids and robotaxis while mitigating industry-wide memory constraints.
• Elon Musk has proactively stockpiled critical electrical infrastructure—such as high-voltage transformers and substations at Giga Texas—to prevent energy grid bottlenecks from stalling compute and mega-pack deployments.

Takeaways

• TSLA offers long-term upside tied to physical AI applications (robotics and autonomous transport) rather than purely digital large language models.
• Watch for execution milestones regarding the TerraFab timeline (targeted within 2 to 3 years) and Cybercab scaling.

SpaceX (Private)

• Leading the private market in rapid deployment of power and data center infrastructure, scaling power capacity from 1.286 gigawatts toward a target of 5 to 10 gigawatts.
• Starship’s reusable launch capabilities and mass-to-orbit capacity are viewed as the primary solution to overcome Earth-bound physical, electrical, and regulatory data center constraints by enabling orbital space-based computing.
• Collaborating with NVIDIA to deploy AI compute directly into space environments via Starlink and StarMind satellites by late 2027.

Takeaways

• Gain exposure to SpaceX via private investment vehicles or secondary markets where possible, as it serves as a critical infrastructure backbone for space-based compute and high-density power delivery.

Palantir Technologies (PLTR)

• Held as a core growth position by guest Hans, citing CEO Alex Karp as one of the top three visionary public market operators alongside Jensen Huang and Elon Musk.
• The company is well-positioned to navigate enterprise and government barriers to entry (such as SOC 2 compliance and security clearances), creating a durable moat and high customer stickiness.

Takeaways

• PLTR remains an attractive software-layer AI play benefiting from high regulatory moats and enterprise compliance requirements.

Rocket Lab USA (RKLB)

• Highlighted as a high-conviction position in the commercial space sector.
• CEO Peter Beck is recognized as an elite operator capable of navigating complex hardware and launch logistics.

Takeaways

• Consider RKLB for diversified exposure to the commercial space and orbital launch ecosystem alongside market leaders like SpaceX.

Sector Theme: AI Infrastructure, Power, & Compute

• The AI infrastructure build-out is projected to last well beyond the previously estimated 3- to 5-year window, driven by near-infinite demand for machine intelligence.
• Power and energy generation represent the single largest bottleneck for data center expansion:
  • Energy constraints are increasing demand for dedicated behind-the-meter generation, high-voltage transformers, and natural gas as a transitional energy source.
• Token economics and software commoditization:
  • Standard reasoning and basic AI token costs have plummeted dramatically (e.g., GPT-3.5 class capability dropping from $20 down to $0.07 per million tokens).
  • Frontier model companies (like OpenAI and Anthropic) face margin pressure from lower-cost open-source and foreign models unless they sustain a distinct reasoning advantage.

Takeaways

• Look beyond chip designers into critical supply chain enablers: electrical equipment (transformers, grid infrastructure), natural gas/energy producers, and semiconductor memory manufacturers.
• In software, favor companies with proprietary distribution, sticky enterprise compliance, or specialized physical applications over undifferentiated wrapper models.