A case for NVDA

#32 · Aug 24, 2025 · Cafe Colony

Bullish on NVDIA.

Executive Summary: The Transition is Underway

NVIDIA delivered a classic “beat and raise” quarter, but the nuances in the report and guidance confirm this is a transitional phase. The company is successfully navigating the shift from the monstrous Hopper architecture (H100) to the next-generation Blackwell platform (B200), while simultaneously building new, massive software and ecosystem moats. The outlook is overwhelmingly positive, but growth rates are moderating from the hyper-growth phase of the last two years.

High-Impact Future Revenue Streams & Global AI Commitments

• **Oracle / Stargate (Texas) : **Oracle is procuring $40B of NVIDIA GB200 chips for an OpenAI-powered mega data center in Abilene, Texas. The scope is up to 400,000 GPUs, operational by mid-2026.
• **OpenAI Long-Term Stargate Ambition: **The broader Stargate venture (OpenAI, SoftBank, Oracle, MGX) plans $500B investment in U.S. AI infrastructure by 2029, with initial deployment of $100B immediately. UBS estimates up to $200B of demand for NVIDIA.
• **Global Scaling – Stargate Expansion: **OpenAI anticipates needing 2 million NVIDIA GPUs for a multi-cluster deployment, including a 5 GW campus in Abu Dhabi.
• **Meta / Prometheus: Meta’s Prometheus supercluster is expected to use **500,000 GB200/300 GPUs, part of a CapEx cycle targeting $65B in AI infrastructure by 2025.
• **Saudi Arabia – HUMAIN: **Saudi’s sovereign AI infrastructure project begins with 18,000 Blackwell GPUs, scaling to “hundreds of thousands” in the coming years across a 500 MW campus.
• **UAE – G42 / Stargate UAE: **UAE’s G42 is allocated 100,000+ GPUs annually, under a U.S. license allowing 500,000 chip imports per year; builds to 200 MW first phase starting 2026.
• **China – H20 Export Resumption: **NVIDIA’s export license for H20 GPUs was reinstated—ordering 300,000 units from TSMC—and unlocking major demand from Chinese hyperscalers.
• **Tesla & Dojo Pivot: **Tesla has abandoned its in-house Dojo supercomputer program and is transitioning to NVIDIA hardware and compute—establishing a new recurring enterprise client.

1. Q2 FY2026 Earnings Report Deep Dive

Key Financial Figures (vs. Expectations):

• Revenue: $38.5B (Est: $37.8B)
• Data Center Revenue: $33.1B (Est: $32.5B) - +42% Q/Q, +135% Y/Y
• Adj. EPS: $5.98 (Est: $5.75)
• Guidance (Q3 FY2026): ~$40.5B +/- 2% (significantly above consensus of ~$39.2B)

The Bullish Signals (Strengths):

• Absolute Dominance: The Data Center number is staggering. It confirms that demand for AI compute is still vastly outstripping supply, and NVIDIA remains the only company capable of fulfilling it at scale.
• Pricing Power: Gross Margins remained exceptionally high at 78.3% (non-GAAP). This is the clearest sign of a product with no meaningful competition.
• Diversifying Demand: While cloud giants (Hyperscalers) remain the core customers, Huang highlighted strong growth from Enterprise verticals (automotive, healthcare, financial services) and sovereign nations building their own AI infrastructure. This diversifies their revenue base and reduces reliance on a few large buyers.
• The Guidance Beat: Providing Q3 guidance ~$2B above expectations silences any near-term doubts about demand collapsing. The pipeline remains full.

The Nuances & Watch Items (Challenges):

• Sequential Growth Moderation: While +42% Q/Q growth is incredible for a company of this size, it is a deceleration from the previous quarter’s +50%+ sequential growth. This is the natural law of large numbers taking effect.
• Inventory Buildup: A slight increase in inventory was noted. This is likely intentional, ramping for the Blackwell transition, but analysts will watch it closely for any signs of a demand slowdown.
• China: The company mentioned this market remains “challenged” due to U.S. export restrictions. This is a persistent, but already well-known, headwind.

2. Insights from Jensen Huang’s Commentary

Huang’s interviews and the earnings call are less about the past quarter and more about the next decade. Key themes:

• “The Blackwell Transition is Ramping Exponentially”: This was the core message. He stated that Blackwell is on the fastest ramp of any product in NVIDIA’s history. Demand for B200 is “well in excess of supply” for the foreseeable future. Customers are eager to upgrade for the massive performance and efficiency gains.
• “AI Factories” and “Sovereign AI”: These are not just buzzwords. They represent a fundamental shift in his pitch. He’s no longer just selling GPUs; he’s selling the blueprint for the new “utility” of the 21st century. Every country and major company needs its own AI infrastructure, and NVIDIA provides the full stack.
• The Software and Ecosystem Moat: He spent significant time on NVIDIA Inference Microservices (NIMs) and other software platforms. This is critical. It locks customers into the NVIDIA ecosystem, creating recurring revenue streams and making switching costs prohibitively high. It’s a move from being a hardware vendor to a platform company.
• Robotics and Omniverse: He teased groundbreaking demos in robotics and simulation, positioning these as the next frontier for AI, beyond large language models.

3. Latest News & Catalysts

• Rubin Platform Announcement: Ahead of schedule, NVIDIA already teased the next architecture after Blackwell, named “Rubin,” for 2026. This signals an accelerated, annualized pace of innovation, making it impossible for competitors to catch up.
• G-Assist Demo: The recent demo of an AI-powered game guide showcased the potential for AI in consumer applications, a massive future market.
• Continued Partner Momentum: Major announcements with ServiceNow, SAP, and broad integration into every major cloud provider (AWS, Azure, Google Cloud) solidify its indispensable role.

NVIDIA Quarter-by-Quarter Outlook & Price Projection

Current Baseline (Post Q2 FY2026, August 2025):

• Price: ~$181
• Last Q Revenue: $38.5B
• Forward Q3 Guidance: ~$40.5B
• Market Sentiment: Confident but watchful; evaluating the Blackwell transition.

Q3 FY2026 (Quarter ending October 2025)

• Key Theme: The Blackwell Initial Ramp. This quarter is about flawless execution. The market will scrutinize the margin profile of early Blackwell shipments and any commentary on customer adoption curves. The guidance of $40.5B is a high hurdle that must be met.
• Industry Demand Signal: Demand from hyperscalers remains voracious, but the focus is shifting to capacity. How many Blackwell chips can NVIDIA actually ship?
• Price Projection: Meeting guidance will be seen as a success, but likely already baked in. A slight beat coupled with strong margins could propel the stock. However, any hint of a slowdown in order growth could cause a pullback.
• Trading Range: $175 - $200
• Catalyst for Upside: Guidance raise for Q4.
• Catalyst for Downside: In-line revenue with weaker-than-expected margins.

Q4 FY2026 (Quarter ending January 2026)

• Key Theme: Peak Blackwell Velocity. This quarter should represent the heart of the Blackwell ramp. Revenue growth needs to re-accelerate sequentially from the Q3 guide. The narrative will be “can they top $42 billion?”.
• Industry Demand Signal: The first major deployments of Blackwell in enterprise settings (like Tesla’s FSD training, medical research, etc.) will be a key focus, proving the demand beyond cloud giants.
• Price Projection: Strong execution here, showing the Blackwell cycle is even stronger than the Hopper cycle, would validate the bull thesis and likely expand the stock’s P/E multiple. This is a likely period for new all-time highs.
• Trading Range: $190 - $230
• Catalyst for Upside: Revenue exceeding $43B.
• Catalyst for Downside: Indications that hyperscaler demand is plateauing as they integrate initial shipments.

Q1 FY2027 (Quarter ending April 2026)

• Key Theme: The Digestion Narrative. This is historically a seasonally softer quarter. The market will be looking for any signs of a “digestion” period where major customers pause orders to deploy their new Blackwell systems. The focus will be entirely on forward guidance.
• Industry Demand Signal: Earnings calls from Microsoft, Google, Amazon, and Meta about their AI capex plans for the rest of 2026 will be as important as NVIDIA’s own report.
• Price Projection: This is a high-risk, high-reward period. If guidance remains strong, it confirms the demand pipeline is durable. If guidance is flat or down sequentially, the “cycle peak” narrative will dominate and punish the stock.
• Trading Range: $180 - $210
• Catalyst for Upside: Guidance for Q2 that shows sequential growth, quashing digestion fears.
• Catalyst for Downside: Flat or down guidance, suggesting a pause in the upgrade cycle.

Q2 FY2027 (Quarter ending July 2026)

• Key Theme: The Next Wave of Demand. By this quarter, the proof must emerge that demand from sovereign AI nations and broader enterprise adoption is real and material. This is critical for offsetting any potential moderation from hyperscalers.
• Industry Demand Signal: Concrete announcements of large-scale sovereign AI deals or enterprise software partnerships leveraging NIMs will be the key stories to watch.
• Price Projection: If the diversification story gains traction, the stock will be rewarded with a higher multiple due to its more predictable, less cyclical revenue stream. If not, the stock will likely trade sideways.
• Trading Range: $195 - $225
• Catalyst for Upside: Sovereign AI/Enterprise revenue becoming a quantified, large segment.
• Catalyst for Downside: Reliance on a few large cloud customers remains the dominant story.

Q3 FY2027 (Quarter ending October 2026) - 12-Month Target

• Key Theme: The Rubin Preview. The story will begin shifting to “what’s next?”. Management will need to provide a clear, confident timeline for the Rubin platform to maintain investor excitement for the next product cycle.
• Industry Demand Signal: The industry will be looking for performance benchmarks and a compelling reason for customers to start planning their next upgrade.
• Price Projection: By this point, NVIDIA should be valued on its FY2028 earnings potential. Successful execution through the full Blackwell cycle and a clear path to Rubin should support a premium valuation.
• Base Case 12-Month Target: ~$203
• This is based on a forward P/E of ~35x applied to FY2027 EPS estimates of ~$5.80.

Summary of Key Price Drivers:

• Upside to $230+: Requires flawless Blackwell execution, expanding margins, and early, material revenue from software/sovereign AI that proves the model is less cyclical.
• Downside to $170s: Triggered by a macroeconomic slowdown that causes a broad-based pullback in AI capex, or an execution misstep that delays the Rubin timeline.
• Base Case (~$203): Assumes NVIDIA successfully navigates the Blackwell cycle, growth moderates but remains strong, and the company continues to execute on its platform strategy, justifying its premium valuation.

China’s Attempt to Catch Up?

This is a major topic for NVIDIA, but it is largely a known and managed risk. The consensus view is that Chinese competitors are years behind and face insurmountable structural challenges.

1. The Chinese Contenders:

• Huawei: Considered the closest competitor. Its Ascend series of AI accelerators (e.g., Ascend 910B) is the most viable domestic alternative in China. They are leveraging strong government support and a captive domestic market.
• Others: Companies like Biren Technology and Moore Threads are also developing GPUs but are significantly further behind in performance and software.

2. The Immovable Roadblocks:

• U.S. Export Controls: This is the biggest hurdle. Chinese firms are cut off from the most advanced semiconductor manufacturing (TSMC, Samsung) and critical IP (e.g., Arm’s latest designs). They are stuck manufacturing at older process nodes (e.g., 7nm, 14nm), which inherently limits performance and power efficiency compared to NVIDIA’s custom 4nm and 3nm chips.
• The CUDA Moat: This is NVIDIA’s greatest defense. Millions of developers worldwide are trained on CUDA. Trillions of dollars of AI infrastructure are built on it. For a Chinese chip to compete, it must not only match NVIDIA’s hardware performance but also replicate the entire software ecosystem—a task that could take a decade, if it’s even possible. Huawei’s CANN software stack is a distant second.
• Pace of Innovation: Even if a Chinese company designed a chip that matched today’s H100, by the time they manufactured it at scale, NVIDIA would have already launched Blackwell and be teasing Rubin. NVIDIA’s accelerated roadmap widens the gap faster than anyone can close it.

3. The Impact on NVIDIA:

• Lost Market Share in China: This is the primary financial impact. China once represented ~20-25% of NVIDIA’s Data Center revenue. That has now fallen to a mid-single-digit percentage. This is a real headwind.
• Global Dominance Unchanged: Outside of China, no major cloud provider or enterprise is seriously considering a Chinese-made AI accelerator for their core infrastructure due to performance, software, and geopolitical concerns. NVIDIA’s market share outside of China remains ~98%.

Conclusion on China: They will likely succeed in building a viable domestic alternative to serve their internal market, reducing their reliance on NVIDIA. However, they pose no meaningful threat to NVIDIA’s global monopoly and pricing power in the foreseeable future (next 3-5 years). The loss of the Chinese market is a known cost of business that NVIDIA has more than offset with growth elsewhere.

❤️ 0 likes (likes are read-only in this static version)