Summary of Key Points
NVIDIA's revenue (96.2 billion US dollars) and profits for the second quarter of its 2027 fiscal year exceeded Wall Street expectations, yet its stock price plummeted by 410 billion US dollars over seven consecutive days. On the surface, the numbers look impressive, but there are three major concerns: the company's revenue is highly dependent on top cloud providers, the financial report contains elements from non-core businesses, and there is a significant increase in cash flow pressure. Jensen Huang is attempting to maintain growth through a combination of strategies including a computing power financing platform, technology acquisitions, and software frameworks, while also facing competition from OpenAI's own chip development.
Detailed Analysis
1. Revenue Exceeds Expectations, but Not Enough to Boost Stock Price
Although NVIDIA's revenue for this quarter surpassed expectations, the main reason for the decline in its stock price is that market expectations have become so high that meeting them is now considered the norm. Additionally, there are underlying issues with the revenue structure:
- High Risk from Single-Engine Dependency: The data center business accounts for 92.5% of total revenue (89 billion US dollars), while other businesses (such as gaming and automotive) combined under "edge computing" only account for 7.2 billion US dollars, indicating nearly stagnant growth in these areas.
- Growth Driven by Key Customers: Large-scale customers (such as Google and Microsoft) contributed 48.7 billion US dollars, 52 billion US dollars more than expected; however, revenue from AI cloud and enterprise customers (ACIE) decreased by 1.7 billion US dollars, suggesting that the demand from smaller customers has not kept up.
- Slowing Growth Momentum: The quarter-on-quarter growth rate for the data center business has slowed from 25% to 18%, indicating a weakening momentum for growth.
- Recovery from Previous Price Surges: NVIDIA's stock price had been rising for over a year before the earnings report, and the 410 billion US dollar drop over seven days is a normal correction of the previous bubble.
2. Hidden Issues in the Financial Report
There are three issues in the financial report that are easily overlooked:
- GAAP Profit Inflation: GAAP (Generally Accepted Accounting Principles) profits are higher than non-GAAP (adjusted actual core business profits) because GAAP includes 7.8 billion US dollars in gains from equity investments (not from chip sales). Non-GAAP profits represent the actual core business earnings and are actually lower.
- Sliding Gross Margin: The gross margin for this quarter was 75%, and it is expected to drop to 73.5%-74.5% in the next quarter, possibly even to 71%-72% in Q4. This is due to increased memory prices (DRAM contract prices have risen by 340%), with profits being taken by memory manufacturers like Samsung and SK Hynix (which have gross margins of nearly 80%).
- Cash Flow Halved: Free cash flow decreased from 48.5 billion US dollars to 21.3 billion US dollars, mainly due to a 160 billion US dollar contract to purchase memory, extended payment terms for customers (increased accounts receivable by 22.3 billion US dollars), and increased capital expenditures of 38-42 billion US dollars.
3. Jensen Huang's Strategies to Break the Pattern
To maintain a market value of 5 trillion US dollars, Jensen Huang is adopting new approaches:
- Computing Power Financing Platform: NVIDIA has partnered with six major asset management firms, including BlackRock and Berkshire Hathaway, to leverage 500 billion US dollars in capital. This allows customers to use future computing power contracts as collateral to borrow money for equipment purchases (similar to how Meituan's riders use electric scooters). NVIDIA also provides 25% residual value support to customers, enabling them to purchase more equipment and thus sell more products.
- Acquiring Technologies Indirectly: For 6 billion US dollars, NVIDIA acquired the model systems from AI programming company Poolside, hiring 109 engineers in the process. This acquisition does not trigger antitrust issues and provides access to core technologies, addressing NVIDIA's shortcomings in large-scale models.
- Software to Bind Customers: NVIDIA has introduced the AVO framework (Agent Execution Framework) to solve issues with long-term AI tasks going awry, making customers dependent on its software and shifting from a focus on hardware sales to a software and hardware integration model.
4. Potential Threat: OpenAI's Self-Developed Chips
Just before the earnings report, OpenAI showcased its self-developed Jalapeño inference chip, which outperforms NVIDIA's Blackwell in terms of throughput per kilowatt. Although OpenAI stated it will continue to use NVIDIA's chips for now, increased production in 2027 could lead to reduced reliance on NVIDIA, posing a long-term competitive threat to the company, which relies heavily on GPU sales.
Conclusion
NVIDIA's current situation is akin to being "on a high platform": its short-term performance is still impressive, but growth momentum is weakening, and profits are being eroded by upstream memory manufacturers. Jensen Huang's new strategies may help maintain growth for the time being, but OpenAI's self-developed chips pose a significant threat. Whether NVIDIA can maintain its 5 trillion US dollar market value in the future will depend on its software development and its ability to compete effectively.