Focus on AI infrastructure over application-layer SaaS, as companies like Mercor are seeing massive growth by providing the high-quality data and human expertise required to train frontier models. Consider long-term positions in "Frontier Labs" like OpenAI and Anthropic, which are projected to reach valuations as high as $10 trillion as they become the essential "teacher models" for the entire industry. While NVIDIA (NVDA) remains a core holding, prepare for a multi-chip future by monitoring emerging competitors like Cerebras and in-house chip projects from Meta and Google. Prioritize investments in AI Security and "agentic trust" platforms like Vanta to capitalize on the urgent corporate need to defend against automated, agent-led cyber attacks. Look for opportunities in the "Token Spend" shift, where enterprise budgets are moving away from human headcount and toward massive compute and inference consumption via platforms like Airwallex.

Mercor

• Mercor is an AI infrastructure company currently valued at $10 billion, reporting over $1 billion in revenue with a 30-40% gross margin. • The company has added $300 million in net new Annual Recurring Revenue (ARR) in just the last 60 days. • Mercor operates a talent network of over 5 million people, paying out over $3 million per day to experts who provide high-quality data to train frontier AI models. • The business is highly profitable and holds over $500 million in cash.

Takeaways

• Infrastructure over Applications: The CEO argues that infrastructure companies (like Mercor) have stronger moats than application-layer companies (SaaS on top of AI), as the "model is the product" and software layers are easily recreated by AI. • Vertical Integration: Mercor is moving from providing raw data to providing "full-stack" tasks, using AI project managers to coordinate human experts. • Market Consolidation: With a large cash reserve, Mercor is positioned to acquire smaller, niche data providers (e.g., medical or legal data specialists) when the market eventually corrects.

OpenAI & Anthropic

• These "Frontier Labs" are viewed as the primary winners in the current AI cycle. • Brendan Foody predicts that at least one of these companies will reach a $10 trillion valuation within five years. • Anthropic is specifically mentioned as a customer of Navan, highlighting its enterprise adoption.

Takeaways

• Teacher Models: Even as open-source models grow, frontier models (OpenAI/Anthropic) remain essential as "teacher models" used to distill and train smaller, more efficient models. • Forward-Deployed Engineering: The defensibility of these labs comes from their "forward-deployed" teams that go deep into enterprises to codify tacit knowledge that isn't written down.

NVIDIA (NVDA)

• Acknowledged as a phenomenal business and a core part of the AI infrastructure. • The CEO notes that while NVIDIA is currently dominant, the industry is moving toward a multi-chip future.

Takeaways

• Chip Diversification: Companies like Cerebras and in-house chip projects from major labs (OpenAI, Meta, Google) are emerging as long-term competitors. • Market Share: Even if NVIDIA’s market share drops to 30-40% due to competition, it is expected to remain one of the most valuable companies in the world due to the sheer size of the AI market.

Investment Theme: The "Token Spend" Shift

• A major shift in corporate spending is occurring: Mercor currently spends more on AI tokens (API calls to models) than on employee headcount. • Prediction: In five years, the average enterprise will spend more on compute/inference than on human salaries.

Takeaways

• Jevons Paradox: As AI models become more efficient and cheaper, total consumption (spend) actually increases because the ROI of using them for more tasks becomes undeniable. • Agentic Finance: Companies like Airwallex are investing in "agentic finance," where AI agents handle banking, treasury, and payments, further driving token consumption.

Investment Theme: AI Security & Cyber Defense

• The transcript highlights a "Golden Age of Cyber" driven by AI-powered attacks. • Attackers are now using swarms of coding agents to find vulnerabilities at a speed humans cannot match.

Takeaways

• Defensive AI: There is an enormous boom expected in AI security engineering tools. • Vanta: Mentioned as a leader in the "agentic trust" space, helping companies manage the risks associated with the rapid adoption of new AI tools.

Sector Insight: The Future of SaaS and Jobs

• SaaS Defensibility: Software companies without a network effect (like Slack Connect or Salesforce’s ecosystem) are at high risk. If a model can "clone Slack" or recreate a workflow via a prompt, the software moat disappears. • Job Displacement vs. Creation: While displacement is happening (e.g., in legal, finance, and consulting), new categories like "Agent Trainers" are emerging. Mercor is paying millions to people to teach agents how to perform complex, multi-week professional tasks.

Takeaways

• High-Value Skills: The most valuable human input is now tacit knowledge—the information in a professional's head that isn't in a manual or on the internet. • AI Researcher Salaries: The cost to hire top-tier AI researchers has skyrocketed, with compensation packages reaching tens of millions of dollars in stock per year.