Monitor Applied Intuition closely for imminent "major announcements" regarding product milestones, as the company leads the shift toward high-value Physical AI in defense and construction. Investors should maintain exposure to NVIDIA (NVDA) due to its massive pricing power with B200/B300 chips, but remain cautious of a potential infrastructure pullback if enterprise ROI fails to materialize. Nebius (NBIS) offers a high-growth opportunity in the data center space as it pivots toward enterprise inferencing and agentic workloads to drive multi-billion dollar revenue. For software exposure, prioritize companies like Glean and Navan that focus on "human-in-the-loop" models and cost-efficient "token budgeting" to ensure actual profitability. Finally, look beyond US-centric tech by exploring "geographic arbitrage" opportunities in Japan and Israel, while favoring data-layer infrastructure like MongoDB (MDB) to capture the essential "Big Data" side of the AI boom.

Applied Intuition

• Applied Intuition is a sizable enterprise AI company with over 1,000 engineers focused on "physical AI"—putting intelligence into machines like cars, trucks, ships, and construction equipment. • The company is working with major partners like Huntington Ingalls (defense/ships) and Heidelberg Materials (mining/ports). • Unlike many AI startups, they have barely touched their funding and are focused on reconciling high model training costs with actual customer revenue. • Upcoming Milestone: The CEO teased the "biggest announcements in the company's history" coming soon regarding product and customer milestones.

Takeaways

• Sector Focus: Look toward "Physical AI" (robotics/autonomous machinery) as a long-term play that may outlast the hype of pure software LLMs. • Stability: The company emphasizes a "linear" growth trajectory rather than the volatile spikes seen in the LLM space, due to the safety-critical nature of physical machines. • Hidden Risk: Investors should adopt the "only the paranoid survive" mindset, looking for hidden risks in technical strategies and leadership even during booms.

NVIDIA (NVDA) / Infrastructure Sector

• Dylan Patel (SemiAnalysis) notes that next-generation hardware (like B200/B300 chips) is seeing price increases even before production starts. • There is a significant risk of "wasted infra spend" because companies are building data centers based on 6-month-old lab data rather than future-proofing for general-purpose flexibility. • Memory Costs: High memory costs are being passed down to customers, preventing the cost of AI tokens from falling as fast as previously expected.

Takeaways

• Pricing Power: NVIDIA and hardware providers currently hold massive pricing power, but there is a looming risk of a "pullback" or a "bubble" reconciliation as companies realize their specific infrastructure might not be optimal for future models. • Infrastructure Strategy: For those investing in the space, flexibility and general-purpose capabilities are more valuable than highly specialized, static hardware setups.

Nebius (NBIS)

• Nebius operates a portfolio of 20 data centers and is expanding rapidly. • They are shifting focus from serving "AI natives" (startups) to driving "enterprise adoption." • The company is moving beyond just model training into inferencing and agentic workloads.

Takeaways

• Diversification: Nebius is positioning itself as a "multi-threaded" player with a diversified portfolio of data center projects to mitigate local regulatory and power risks. • Revenue Growth: The company expects to generate tens of billions in revenue as it scales into big-brand enterprise adoption over the next year.

Glean

• Glean is a leader in "context graphs," helping enterprises use their own internal data to power AI agents. • They focus on ROI by helping companies reduce "token usage" (costs) by picking the right model (open source vs. closed) for specific tasks.

Takeaways

• Cost Efficiency: As enterprises become more cost-conscious, companies that provide "context" and help manage "token budgets" will likely see higher demand. • Open Source Dominance: There is a strong belief that open-source models will dominate AI inferencing within the next two years.

Navan

• Navan (formerly TripActions) is a travel AI company that recently became cash-flow positive and profitable. • They handle over $10 billion in bookings annually and grew revenue by 40% last quarter. • Human-in-the-loop: The CEO emphasizes that for high-stakes industries like travel, AI cannot function alone because "hallucinations" (errors) are unacceptable; human support remains a critical component.

Takeaways

• Profitability Matters: Navan serves as a case study for AI companies successfully transitioning from high growth to actual profitability. • Hybrid Models: Investment opportunities may be strongest in companies that use AI to augment humans rather than trying to replace them entirely in complex, emotional sectors like travel.

Investment Themes & Sector Trends

1. The "Four Seasons" of AI

• Apoorv Agrawal (Altimeter) describes the current landscape as having four "seasons": OpenAI, Anthropic, SpaceX (xAI), and Google. • Insight: Investors and enterprises should plan for the "climate" (long-term multi-model strategy) rather than the "weather" (which specific model is winning this week).

2. Open Source vs. Closed Models

• There is a conflicting view: some believe open source will dominate inferencing (Glean), while others (SemiAnalysis) suggest that "Open is dying" as labs move toward licensing models to recoup costs. • Insight: Watch the "token maxing" trend. Some experts believe "token budgeting" is a "loser mentality" that prevents companies from evolving, while others warn that excessive token spend without ROI is a bubble.

3. Geographic Arbitrage

• Wonderful AI suggests that "geographies will be bigger than verticals." • Insight: There is a massive "land grab" opportunity in providing AI solutions to the "rest of the world" (outside the US), specifically in markets like Japan and Israel.

4. The Return of Big Data

• MongoDB (MDB) and Turbo Puffer representatives emphasize that "Data is back." • Insight: An AI application is only as good as its data layer. Companies providing "vector search" and "object storage" (like S3 or Google Cloud Storage) at scale are the "unsung heroes" of the current boom.

5. Risk of an "LTCM" Moment

• Multiple speakers warned of a potential "reconciliation" or "bust." • Risk Factor: A comparison was made to Long-Term Capital Management (LTCM), where "geniuses" used too much leverage and failed. In the current context, the "leverage" is massive compute spend and high valuations without immediate revenue.