The global investment community's attention turns to NVIDIA, the "AI bellwether," which will release its latest earnings report after U.S. market close tomorrow (Thursday morning Beijing time). This earnings report serves not only as NVIDIA's report card but also as a "litmus test" for the global artificial intelligence wave, with influence sufficient to sway sentiment across the technology sector and the broader market. How high are market expectations? Can NVIDIA once again exceed the ceiling? Which metrics will be the real "game-changers"? Facing significant volatility risks, how should investors position themselves?

**Business Outlook**

When NVIDIA reports its second-quarter results Wednesday afternoon, it will undoubtedly deliver some impressive numbers. Whether these numbers are impressive enough is an entirely different question. Here are the expectations for the quarterly report figures:

Total Revenue: Expected at $46 billion, representing a 53% increase from $30 billion in the same period last year, and above the previous quarter's guidance of $45 billion. Earnings per share are expected at $1.01.

Data Center Revenue: Expected at $41.3 billion, up 57% year-over-year. While this growth rate is robust, it falls short of the 73% growth reported in NVIDIA's previous earnings report.

Profitability: Adjusted operating profit is expected at $29.1 billion. Net profit is anticipated at $24.7 billion, up from $16.6 billion in the same period last year.

Revenue Guidance: $53.4 billion for the third quarter ending in October.

For the second quarter ending in July, the chip manufacturer's revenue and adjusted operating profit are expected to reach new highs, driven by continued surge in artificial intelligence (AI) system spending. For NVIDIA shareholders, this represents another boom period, as the company's stock has weathered recent selling pressure that affected other AI-related technology stocks.

NVIDIA's stock price has risen nearly 39% over the past three months, significantly outperforming other large technology stocks during the same period. Moreover, the company's $4.4 trillion market capitalization is currently 20% higher than Microsoft's.

During the last earnings release, the company excluded its China business from current quarter guidance, and many analysts expect the company to continue this approach. However, FactSet data shows that since H20-related news emerged last month, market expectations for NVIDIA's data center revenue for the quarter ending in October have been revised up by 5% on average. This could make it more difficult for the company's guidance to exceed Wall Street targets, something the company has surprisingly accomplished frequently over the past three years.

**Investor Focus Points:**

Piper Sandler's Harsh Kumar states that NVIDIA's guidance around its China business will be the focus of Wednesday's earnings report.

On the afternoon of August 21st local time, technology media The Information, citing informed sources, reported that NVIDIA had instructed key component suppliers including South Korea's Samsung Electronics and U.S.-based Amkor Technology to suspend production related to H20 chips. Additionally, Reuters, citing informed sources, confirmed that NVIDIA had asked Foxconn to suspend H20 chip production. In response to these reports, NVIDIA stated: "We continuously manage our supply chain to respond to market conditions," but declined to provide further details.

However, analysts still see upside potential for H20 chip sales. JPMorgan analysts wrote in a Monday report that NVIDIA's existing H20 inventory could generate up to $6 billion in incremental revenue. Kumar indicates that the China business could achieve double-digit growth in the coming quarters.

**Supply and Demand Dynamics**

For several quarters, NVIDIA chips have been in short supply, and Wall Street will watch for signs of continued strong demand to gauge the company's potential for future earnings beats and the growth trajectory of the data center industry.

Naji observed an increase in announced 2026 data center projects, which could bring NVIDIA up to $100 billion in incremental revenue over the next year. Demand is increasingly coming from sovereign clouds and new cloud developments outside the United States, indicating NVIDIA's business breadth is expanding.

The company's Blackwell chips have achieved significant production increases, with next-generation Rubin chips planned for launch in the second half of next year. Cantor Fitzgerald's C.J. Muse will focus on more specific demand forecasts Wednesday. He states that NVIDIA CEO Jensen Huang should have "clearer expectations" for data center demand conditions entering calendar year 2026, including the capacity ramp-up outlook for Rubin chips expected to launch later in calendar year 2026.

**Will Hyperscale Cloud Providers Continue Spending?**

Hyperscale cloud providers - the companies building AI infrastructure - have been continuously raising their capital expenditure forecasts and purchasing NVIDIA's graphics processing units (GPUs) in large quantities. These companies' multi-billion-dollar budgets represent a major positive for NVIDIA. However, last week's tech-led stock market decline left some investors uneasy about the sustainability of these massive investments.

Additionally, in NVIDIA's first-quarter report, the company revealed that 30% of revenue came from just two customers, indicating high customer concentration. Kumar remains optimistic about hyperscale cloud provider spending, believing these companies are "engaged in a race around artificial general intelligence (AGI) services" and "have plans to maintain such spending levels, which we believe should continue for several more years."

**Potential for Significant Stock Volatility**

U.S. options market data shows that options traders expect NVIDIA's market capitalization could fluctuate by approximately $260 billion after the chip manufacturer releases its latest quarterly report Wednesday after market close (Thursday morning Beijing time). Traders are closely watching NVIDIA's earnings to determine whether its $4 trillion market valuation is justified. Investment rating data shows that over the past six months, 11 institutions have given NVIDIA buy ratings, with an additional 13 providing recommendations. Since July, institutional target prices have ranged from a high of $206 to a low of $190.

**The "New Brain" for Robots**

On August 25th, NVIDIA announced that Jetson Thor, based on the Blackwell architecture, is now officially available to developers. This is a supercomputer designed for robots that can provide computing power support for millions of robots across manufacturing, logistics, transportation, healthcare, agriculture, and retail industries.

According to the introduction, compared to the previous generation Jetson Orin, Jetson Thor delivers 7.5 times the AI computing power and 3.5 times the energy efficiency. It can run various generative AI models, including general models like Cosmos Reason, DeepSeek, Llama, Gemini, and Qwen, as well as robot-specific models like Isaac GR00T N1.5. Currently, the Jetson Thor developer kit is available for $3,499, with Jetson T5000 modules priced at $2,999 per unit for purchases of 1,000 or more.

From computers and data centers to robots, NVIDIA's GPU computing platform has reached more endpoints and accelerated expansion. In recent years, robotics has become an important business line for NVIDIA, often appearing as the finale at company presentations. Jetson Thor can be considered the current "most powerful brain" for robots, capable of integrating "large models + real-time sensing + control" at the edge. Based on Blackwell GPU and 128GB memory, Jetson Thor delivers up to 2,070 FP4 TFLOPS of AI computing power, enabling parallel execution of multimodal models like vision-language-action (VLA) on the robot itself, significantly reducing cloud dependency and round-trip latency.

This also means that after being equipped with Jetson Thor, robots will achieve even higher levels of intelligence, efficiently handling complex application scenarios such as agentic AI, high-speed sensor data processing, and general robotics tasks.