Nvidia Q3 earnings: Why the setup for Nvidia is looking very good 'from multiple angles'

Key Concepts

• Booked Business: Pre-committed orders for future products and services.
• Blackwell and Rubin: Nvidia's next-generation GPU architectures.
• Sequential Trajectory: The quarter-over-quarter growth pattern of a company's revenue.
• Data Center Revenues: Revenue generated from sales of hardware and services for data centers.
• Consensus Expectations: The average forecast of financial analysts regarding a company's performance.
• Capex Plans (Capital Expenditure): Investments made by companies in physical assets like buildings, machinery, and technology.
• Hyperscalers: Large cloud computing providers such as Amazon, Google, Microsoft, and Meta.
• Supply Chain Bottlenecks: Constraints or limitations in the production or delivery of goods.
• HBM Memory (High Bandwidth Memory): A type of high-performance RAM used in GPUs.
• Co-packaged Optics: Integration of optical components directly onto the same package as the semiconductor chip.
• AI Shells: Infrastructure designed to house and power AI computing hardware.
• Gigawatts: A unit of power, referring to the electrical energy required for data centers.
• Earnings Power: A company's ability to generate profits.
• Replatforming General Theory: A framework for understanding the economy's capacity to absorb technological infrastructure investments.
• Productivity Gains: Improvements in efficiency and output.
• Margins Compression: A decrease in profit margins.
• AI Infrastructure Stack: The entire ecosystem of hardware, software, and services required for artificial intelligence.

Nvidia's Strong Outlook and Market Position

Antoine Skybond, a street research technology infrastructure analyst, discusses Nvidia's exceptionally strong position and future prospects, driven by significant booked business and robust customer investment plans.

1. Booked Business and Revenue Projections

• Jensen Huang's Comments: Nvidia CEO Jensen Huang indicated at the GTC event in Washington D.C. that the company has secured "booked business" for its Blackwell and Rubin architectures, covering 2025 and 2026.
• Cumulative Booked Business: This booked business amounts to a substantial $500 billion.
• Implied Sequential Growth: This forecast implies that Nvidia's data center revenues will grow by approximately 20% sequentially each quarter through the end of 2026.
• Discrepancy with Consensus: Current consensus expectations among analysts do not fully reflect this projected upside.

2. Customer Investment Alignment

• Top-Down Perspective: Analyzing the capital expenditure (capex) plans of major customers like Amazon, Google, Microsoft, and Meta reveals a strong alignment with Nvidia's projected growth trajectory.
• Consistency: These hyperscalers' investment plans are "very consistent" with the implied sequential growth for Nvidia.
• Positive Setup: This convergence of factors suggests a very strong setup for Nvidia from multiple angles for the upcoming quarters and into 2026.

Supply Chain Considerations

Despite high demand, Skybond expresses limited concern regarding potential supply bottlenecks for Nvidia.

1. Manufacturing and Component Supply

• Potential Bottlenecks: Bottlenecks could arise in the manufacturing of GPUs themselves, particularly concerning High Bandwidth Memory (HBM) and co-packaged optics required for chip assembly.
• Nvidia's Efforts: Nvidia has "worked very hard to secure supply," leading Skybond to believe these specific areas are unlikely to be major bottlenecks.

2. AI Shell Rollout and Power

• CoreWeave's Issues: Some companies, like CoreWeave, have reported issues in rolling out powered AI shells.
• Scale and Diversification: Skybond suggests these issues are likely isolated and related to the scale of the company. Larger customers, with more diversified partners, are better positioned to manage these rollouts.
• Power Availability: Concerns about actual power (gigawatts) availability are also deemed unlikely to be a significant issue due to potential efficiencies found throughout the supply chain.
• Overall Supply Outlook: Skybond is "not too concerned about supply being an issue in 2026."

Stock Performance and Valuation

The discussion shifts to Nvidia's stock performance, acknowledging its recent gains but also its historical tendency to decline post-earnings reports, even on beats.

1. High Expectations and Stock Reaction

• Recent Gains: The stock is up approximately 40% year-to-date.
• Historical Pattern: Despite consistent earnings beats, the stock has declined in five of the past eight earnings reports.
• High Expectations: Current expectations are very high, leading to questions about whether the stock will perform in line with potential earnings beats.

2. Earnings Power and Future Valuation

• Projected Earnings: If Nvidia delivers on its booked business, its earnings per share (EPS) could approach $9 next year.
• Current Trading Multiple: The stock is currently trading at a low 20s multiple.
• Forward-Looking Valuation: As the company moves into 2026, the stock will begin trading on 2027 numbers.
• Customer Commitments: Commitments from major customers, such as Amazon planning to double its data center footprint by 2027 and similar comments from Microsoft, provide visibility.
• 2026 Not the Peak: These indicators suggest that 2026 is unlikely to be the peak for Nvidia's growth.
• Stock Reflection: Skybond believes these strong future numbers will eventually be reflected in the stock price, leading to a "very, very strong setup for Nvidia as we enter next year."

The "Bubble" Question and Long-Term Outlook

The conversation addresses the potential for a market bubble and the long-term sustainability of Nvidia's growth.

1. When is the Peak?

• Bubble Uncertainty: Skybond acknowledges that predicting the exact timing of a market bubble bursting is impossible.
• Near-Term Visibility: Strong near-term signals exist, with 2026 being "really well set" and good visibility for 2027.

2. Economic Absorption Capacity

• Replatforming General Theory: Skybond's "replatforming general theory" suggests the global economy has significant capacity to absorb data center capex.
• Absorption Potential: The economy can absorb "north of $3 trillion" in data center capex, compared to the current half a trillion for the top four hyperscalers next year.
• Productivity Gains: As long as the economy can extract productivity gains and continue to grow with this infrastructure, there's no inherent reason to stop investing.

3. Risks and Long-Term Scenarios

• Potential Pause: A pause in investment could occur if a hyperscaler attempts to "replatform too quickly," investing excessively in infrastructure, cash costs, or workforce.
• Margin Compression: Such rapid investment could lead to margin compression and a slowdown in investment.
• Long-Term Growth Scenario: A scenario exists where global data center spending could grow "10x" between now and 2035, indicating substantial room for growth beyond any potential bubble concerns.

Investor Actions and Market Sentiment

The discussion touches upon recent significant investor actions and their implications.

1. SoftBank and Peter Thiel's Sales

• SoftBank's Sale: SoftBank sold its Nvidia shares to invest in OpenAI.
• Peter Thiel's Sale: Peter Thiel also announced the sale of his Nvidia shares.
• Interpretation: Skybond views these actions not as a cause for concern but as investors seeking upside in different parts of the AI infrastructure stack, from chips to models and applications. Different investors may have varying views on where the most value will be captured.