Nvidia's Investing Strategy Is 'Smart', Says T. Rowe Price's Wang

Key Concepts

• Agentic AI: AI systems capable of autonomous, persistent task completion over extended periods rather than simple, one-off queries.
• Scaling Laws: The principle that frontier AI models improve in performance and efficiency as more compute power is applied.
• Hyperscale vs. Enterprise/Sovereign Demand: The distinction between massive cloud providers (hyperscalers) and the emerging, broader adoption of AI by corporations and governments.
• Inference vs. Training: The shift from building models (training) to running them for real-world applications (inference), which is expected to be a larger market.
• PEG Ratio (Price/Earnings-to-Growth): A valuation metric used to determine if a stock is undervalued relative to its growth rate.
• Silicon Photonics: A technology used to increase data transmission speeds and efficiency within data centers.

---

1. Market Dynamics and Nvidia’s Growth

Tony Wong, Portfolio Manager at T. Rowe Price, argues that Nvidia’s growth is "unprecedented" and that traditional semiconductor investment playbooks—which suggest selling when margins peak—are failing to account for the unique nature of current AI demand.

• The "Wall of Worry": Investors are concerned about the sustainability of growth, but Wong contends that end-demand is robust due to the transition toward Agentic AI.
• Demand vs. Supply: Jensen Huang (CEO of Nvidia) noted that while supply chains are doubling or quadrupling annually, global demand still vastly outstrips capacity. Wong views this supply-constrained environment as an "enviable position" rather than a weakness.

2. Expanding the Total Addressable Market (TAM)

A key argument presented is that Nvidia is no longer solely dependent on hyperscale capital expenditure (capex).

• Diversification: Nvidia is successfully expanding into enterprise adoption (notably in financial services) and robotics.
• Revenue Sustainability: By breaking out hyperscale versus enterprise/sovereign demand, Nvidia demonstrates that its growth is not just a "hyperscale bubble" but a broader technological shift.

3. Strategic Capital Allocation

Nvidia’s strategy of investing in its ecosystem is highlighted as a major competitive advantage.

• Ecosystem Building: Nvidia uses free cash flow to make small, strategic investments ($2B increments) in private market partners. This secures the supply chain and ensures the technology frontier continues to advance.
• Capital Return Program: Wong draws a parallel to Apple’s historical capital return program. He suggests that as Nvidia matures, consistent capital returns will make the stock less cyclical and lead to multiple expansion, even if the market doesn't react immediately.

4. Technical Arguments: Scaling Laws and Inference

• Efficiency of Frontier Models: Wong argues that using the most advanced models is actually cost-effective. Because these models are smarter, they avoid the "rabbit holes" of less capable models, ultimately saving money on compute.
• Inference Dominance: While training remains a growth driver, Wong emphasizes that inference (the actual usage of AI) will eventually become a much larger market than training.
• Hardware Longevity: Addressing the "bear case" regarding GPU obsolescence, Wong notes that the 3–4 year warranty and depreciation cycles are standard, and even older hardware (like A100s) remains highly valuable and in use, contradicting fears of rapid hardware irrelevance.

5. Valuation and Outlook

• Attractive Multiples: Nvidia is currently trading at approximately 22x forward 12-month earnings, which is significantly lower than its historical average of 34x.
• Growth Durability: Given the current PEG ratio and the shift toward inference, Wong believes there is significant room for multiple expansion.
• The "Jensen Equation": The core thesis remains: More compute = More tokens = More revenue. Wong confirms that this is playing out in real-world cloud demand and enterprise AI development.

---

Synthesis and Conclusion

The core takeaway is that Nvidia has successfully transitioned from a hardware component supplier to a foundational single-architecture platform. The market’s skepticism regarding the sustainability of AI growth is being countered by the shift toward Agentic AI and the massive, untapped potential of enterprise and sovereign adoption. By maintaining a disciplined capital return program and aggressively building out its ecosystem, Nvidia is positioning itself to move from a cyclical semiconductor play to a durable, long-term platform company. The primary indicator of continued success will be the sustained demand for inference, which is expected to dwarf the initial training phase of the AI revolution.