NVIDIA’s Q3 Earnings Reveal a Structural Turning Point in the AI Semiconductor Market
Editor
New Growth Drivers in the AI Semiconductor Market: Visibility from Blackwell and Rubin
The figures revealed by NVIDIA, headquartered in Santa Clara, California, during its Q3 2025 earnings call, clearly illustrate the structural changes in the AI semiconductor market. The company announced that it has secured $500 billion in revenue visibility from its next-generation Blackwell and Rubin chip platforms. This is not just a performance metric for a single company but a key indicator of how rapidly the entire AI infrastructure market is growing. The fact that demand for AI infrastructure consistently exceeds expectations sends an important signal to market participants.
Notably, the Blackwell GB300 model accounts for about two-thirds of total Blackwell revenue, highlighting the strong demand for premium products designed for high-performance AI workloads. NVIDIA’s projection that half of its long-term business opportunities will arise from hyperscalers’ transition to accelerated computing and generative AI suggests that major cloud service providers like Amazon, Google, and Microsoft will continue to expand their investments in AI infrastructure. Indeed, NVIDIA announced its collaboration with OpenAI to build and deploy AI data centers exceeding 10 gigawatts, an unprecedented scale for a single project.
The Rubin platform is expected to enter full-scale production in the second half of 2026, demonstrating how systematically NVIDIA’s technology roadmap is planned. The fact that the A100 GPU, shipped six years ago, is still being produced at maximum capacity today indicates that the lifecycle of AI chips differs from that of traditional semiconductors. This suggests that AI workloads require sustained computing power and that once established, infrastructure is utilized over the long term.
Geopolitical Risks and the Complexity of the Chinese Market
One of the most notable aspects of NVIDIA’s performance is the impact of geopolitical issues. The company stated that a large order for the Hopper platform anticipated in Q3 did not materialize due to geopolitical issues with China. The fact that Q3 sales of the H20 AI chip, specifically designed for the Chinese market, amounted to only $50 million reflects this situation. More importantly, significant purchase orders for the H20 chip were not realized due to geopolitical issues and China’s increasingly competitive market environment.
This situation highlights the complexity of the global AI semiconductor market. China is one of the world’s largest semiconductor consumption markets, but the US-China tech supremacy competition restricts the export of advanced AI chips. While NVIDIA expressed disappointment over trade sanctions with China, its commitment to ongoing collaboration with government authorities is interpreted as an effort to manage these geopolitical risks. Meanwhile, in China, local companies like Baidu and Alibaba are accelerating their own AI chip development, intensifying competition.
For Korean memory semiconductor companies like Samsung Electronics and SK Hynix, this situation can be a double-edged sword. High-bandwidth memory (HBM) is essential for enhancing AI chip performance, and as constraints in the Chinese market grow, opportunities for Korean companies may increase, but there is also the risk of an overall market size reduction. Particularly, SK Hynix, as a key HBM supplier for NVIDIA, is in a position where it must respond sensitively to geopolitical changes.
NVIDIA’s forecast that Physical AI will be the company’s next growth driver and a trillion-dollar opportunity suggests that demand for AI semiconductors in fields like robotics, autonomous driving, and smart manufacturing will explode. This implies that AI systems operating in real physical environments will create new markets, extending beyond data centers or cloud computing. Tesla’s FSD (Full Self-Driving) chip and Boston Dynamics’ robots can be considered early examples of this Physical AI.
According to NVIDIA’s CFO, the company plans to expand its manufacturing capacity in the US over the next four years in collaboration with Taiwan’s Siliconware Precision Industries and America’s AMKOR Technology. This is interpreted as a strategic move to diversify the supply chain and mitigate geopolitical risks. Especially as it aligns with US government semiconductor manufacturing promotion policies like the CHIPS Act, it is expected to accelerate the restructuring of the global semiconductor manufacturing ecosystem.
From a financial perspective, NVIDIA announced plans to maintain its gross margin in the mid-70% range during the fiscal year 2027. This is an indicator of strong pricing power in the AI semiconductor market. The announcement of an expected 14% quarter-over-quarter growth in Q4, driven by Blackwell momentum, suggests that short-term growth momentum will continue. However, the explicit exclusion of China’s data center computing revenue from Q4 guidance indicates that geopolitical uncertainties still exist.
CEO Jensen Huang’s response to the question “Is this an AI bubble?” provides important insights into understanding the nature of the current AI market. He emphasized that the underlying investment frenzy is rooted in a structural change in the computing paradigm. His core argument is that this is not a speculative bubble but a demand based on technological inevitability. The first pillar he presented is the transition to accelerated computing following the end of Moore’s Law. While CPU-based general-purpose computing has reached its speed improvement limits, the demand for computing is exploding, making GPU-based accelerated computing essential.
In fact, six years ago, 90% of supercomputers were CPU-based, but now that figure has plummeted to 10-15%, while GPU/AI accelerated computing has reversed from 10% to 90%, according to Huang. This is clear evidence that the global computing infrastructure is undergoing a major shift from CPU to GPU. The second pillar is the AI transition of the existing internet engine, the recommendation system (RecSys). As the recommendation system, which has been the core of the internet for the past 15 years, evolves into a generative AI method, it is shifting from CPU-based to GPU-based. The third pillar is the emergence of Agentic AI. As entities like OpenAI, Anthropic, and Google’s Gemini expand beyond simple tools to become agents that make decisions and act independently, GPU demand is skyrocketing.
This analysis suggests that the growth of the current AI semiconductor market is not a temporary phenomenon but is due to a fundamental change in computing architecture. While traditional CPU-centric companies like AMD and Intel are making significant investments in developing GPUs and AI accelerators to respond to this change, NVIDIA’s CUDA ecosystem and software stack remain robust. Especially with Taiwan’s TSMC providing cutting-edge process technology as NVIDIA’s key foundry partner, this technological advantage is likely to persist for the foreseeable future.
In conclusion, NVIDIA’s Q3 earnings announcement shows that the AI semiconductor market has moved beyond a simple growth stage and entered a structural transition period. The $500 billion revenue visibility, expansion into Physical AI, geopolitical risk management, and fundamental changes in the computing paradigm all suggest that this market has sustainable growth drivers in the long term. However, geopolitical conflicts with China, the pursuit of competitors, and supply chain stability challenges still exist, requiring cautious strategic approaches from market participants.
**Disclaimer**: This article is for informational purposes only and is not intended as investment solicitation or advice. Investment decisions should be made based on individual judgment and responsibility, and the stock prices or financial performance of the mentioned companies are not guaranteed. All investments carry risks, so please make careful decisions through thorough research and expert consultation.
