Doug Clinton: Nvidia Still On Top with "Using AI"

Key Concepts

• NVIDIA's Revenue and Profitability: Strong performance driven by AI adoption.
• Blackwell Chips: NVIDIA's new chip architecture.
• Hyperscalers' CapEx: Increased capital expenditure by major cloud providers (Meta, Google, OpenAI) to meet AI demand.
• China Market: Potential for NVIDIA to re-enter the Chinese market with modified chips.
• Export Controls: US government restrictions on selling advanced chips to China.
• H20 Chips: NVIDIA chips previously sold in China, now restricted.
• GB200 (Grace Blackwell): NVIDIA's data center chips.
• Inference vs. Training: The difference between building AI models (training) and using them (inference).
• Data Centers: Facilities housing the computing infrastructure for AI.
• AI Adoption: The increasing use of AI across various industries.

NVIDIA's Financial Performance and Future Growth

The discussion begins by highlighting NVIDIA's "phenomenal revenue growth" and "phenomenal profitability." The release of new Blackwell chips and the broader adoption of AI are seen as key drivers for future growth. The speaker mentions using AI for portfolio management, indicating a strong belief in AI's transformative potential.

Hyperscalers and Capital Expenditure

All major hyperscalers (Meta, Google, OpenAI) have indicated they are "continuing to spend on CapEx" and increasing their CapEx guidance. This suggests strong and sustained demand for NVIDIA's products, particularly for AI-related infrastructure. This increased spending instills confidence that NVIDIA's upcoming numbers will be strong.

China Market and Export Controls

A significant portion of the discussion revolves around NVIDIA's potential to re-enter the Chinese market. Reports suggest NVIDIA has developed alternative chips to comply with US export controls. The key question is whether the US administration will "lighten the export controls" to allow NVIDIA to sell these chips in China.

When NVIDIA was disallowed from selling its H20 chips, they wrote down $5.5 billion worth of those chips. This represented an estimated $15 billion annualized hit to revenue. Re-entering the Chinese market, even in a limited capacity, could potentially unlock a $10 billion upside. The speaker predicts NVIDIA will find a way back into China within a couple of months.

The $15 billion figure is identified as the 2026 sales projection hit due to H20 restrictions.

Blackwell Architecture and GB200

NVIDIA is adapting its Blackwell architecture for the Chinese market, potentially creating a "cutout version" to replace the H20. This version may be lower-powered than the chips sold in the US. The GB200 (Grace Blackwell) chips, designed for data centers, will likely be different from the chips sold in China. Reports suggest a potential sales price of $6,000-$8,000 for the Chinese market chips, compared to $40,000 for chips in the US.

Sustainability of NVIDIA's Growth

The conversation addresses concerns about how long NVIDIA can sustain its rapid growth, given the "relentless demand" for its chips. While the addressable market is vast, there is an eventual limit. The speaker believes the growth can last longer than many investors anticipate. Street numbers already factor in a slowdown to "mid-teens kind of growth" next year, compared to over 100% growth in the past year.

Hyperscalers are reporting "off the charts" demand and an inability to fully serve customer needs. This suggests continued investment in new data centers, particularly those focused on inference, which could provide a longer tailwind for NVIDIA.

Inference vs. Training and Revenue Implications

The discussion explores the relationship between the computing power required for training AI models versus inference. While training is a "discrete time bound activity," inference represents an "ongoing and persistently growing demand."

The speaker does not anticipate a drop in revenue. Instead, previously used training resources can be repurposed for inference. Dedicated inference stacks and data centers may also be built. The overall demand for additional compute is expected to remain high.

Conclusion

NVIDIA's strong financial performance is driven by the widespread adoption of AI. The company faces challenges and opportunities in the Chinese market due to export controls. While growth is expected to moderate, sustained demand from hyperscalers and the increasing importance of AI inference suggest a longer-term positive outlook for NVIDIA.