The news agency Xinhua's report on the NVIDIA investigation is a major counterattack in the Tech War against China.
A few days ago, Xinhua News Agency performed a rare task by pointing its sights on NVIDIA’s antitrust violations. Thus, it is the first time Xinhua agency addressed such investigation at the filing step. Traditionally, inquiries into companies like Alibaba and Tencent were only made by the media after the final verdict. What is the difference between this scenario and the rest?
This suggests that the investigation into NVIDIA is a highly targeted action.
NVIDIA: The Cornerstone of American Hi-Tech Global Power
At the moment, the technology sector is the most prominent pillar of the US economy, and its underpinning principles are the “Magnificent 7 Stocks” tech corporations, which are:
Microsoft, Apple, NVIDIA, Alphabet (Google’s parent company), Amazon, Meta, and Tesla.
With seven representatives of those companies, by the end of 2023, they contributed approximately a quarter of the total market value of the S&P 500 index and accounted for two-thirds of the index’s growth. Their aggregate value easily makes them the most valuable businesses in the world—surpassing the total value of all Japanese, French, and British companies combined.
Moreover, a big player among them was NVIDIA, which became the world’s most valuable company with its market capitalization reaching $3.29 trillion, surpassing the $2 trillion mark. The respectful mighty cannot fall down because the impact of collapse will be comminuted through the U.S. technology sector and the economy as a whole.
NVIDIA? What Are This Investigation’s Peculiarities?
Regularly, antitrust actions finish with enormous financial penalties designed to limit the monopolistic greed. Nevertheless, the task of NVIDIA includes its agreement in respect of the Israeli technology business Mellanox.
The Chinese response to NVIDIA’s acquiring of Mellanox was a commitment of NVIDIA towards providing those products and technologies without discrimination. Yet, due to many geopolitical facts, NVIDIA has now renounced to offer critical artificial intelligence chips and corresponding technologies to China and is in contradiction to the former parity request. NVIDIA maintains its proclivity for market leadership as a function of development innovation, while the main issue with the Chinese market is not the NVIDIA advantages, but its failure to deliver on promises.
This investigation underscores a crucial point: China’s technological advancements have made NVIDIA’s blockade ineffective.
The next phase of the AI’s development will indeed need the highest level of application of computing power, as industries become affected more and more by this gigantic technology. The progress of AI models in this perspective is an issue rather than just raw computing power—it’s more and more dependent on data transmission. Our traditional electronic communication methods, like PCIe and InfiniBand, have been a hurdle that we need to double-check at the growth in AI task workload. Because of these problems, optical communication has been elevated to the status of a cutting-edge technology that can overcome the limits of the speed of data transmission, latency, as well as energy efficiency.
Optical Communication: The Key to AI Computing Power
Optical communication technology plays a decisive role in enhancing AI computing power, particularly in terms of bandwidth, latency, and energy efficiency. As artificial intelligence advances rapidly—especially with large-scale AI models like GPT-4, Gemini, and Claude—traditional electronic communication is becoming a computational bottleneck. Optical communication offers a high-speed, low-energy, and stable alternative.
1. The Bottleneck in AI Computing Growth: Data Transmission vs. Processing Power
A major challenge in AI development today is the following contradiction:
-
Computing power is advancing rapidly – AI chips like NVIDIA’s H100 and Huawei’s Ascend 910B continue to evolve, while specialized hardware such as FPGA and ASIC keeps improving.
-
Data transmission rates are lagging behind – The increasing computational power leads to massive data exchange demands, but traditional electronic transmission methods (such as PCIe and InfiniBand) have limited bandwidth, creating a bottleneck for AI expansion.
For example, in large-scale AI training, GPUs must frequently exchange data. High latency can cause processing units to wait for data, reducing overall computing efficiency. This is where optical communication becomes a game-changer.
2. How Optical Communication Enhances AI Computing
Compared to traditional electronic communication, optical communication offers three major advantages:
-
Ultra-high bandwidth:
-
Traditional InfiniBand can reach a maximum of 800Gbps.
-
Optical communication systems, such as Huawei’s Xinghe 16800, can achieve 9,642Tbps, nearly 200 times faster than NVIDIA’s solution.
-
This enables AI computing centers to connect multiple GPUs, TPUs, and NPUs more efficiently, preventing bandwidth limitations from affecting performance.
-
-
Ultra-low latency:
-
Optical signals travel at nearly the speed of light, while electronic signals suffer from severe copper wire losses, requiring multiple relay amplifications.
-
Optical communication eliminates long-distance transmission delays, enhancing AI training and inference speeds, especially for large-scale AI clusters such as Microsoft Azure AI and Google TPU Cloud.
-
-
Minimal energy consumption:
-
Electronic transmission suffers from severe signal loss at high speeds, requiring additional power amplifiers, increasing energy costs.
-
Optical communication consumes far less energy, significantly reducing data center power usage and carbon emissions.
-
3. Optical Communication in AI Computing Centers
Global tech giants have already widely adopted optical communication in AI data centers:
-
NVIDIA: The DGX GH200 AI supercomputing system integrates InfiniBand and optical communication for improved GPU interconnectivity.
- Huawei: The Xinghe 16800 switch, based on optical communication, provides 9.6Pbps bandwidth, vastly outperforming NVIDIA’s solutions.
-
Google: TPU v5 clusters use optical fiber interconnects to minimize computational loss and improve training efficiency.
As photonic computing technology advances, optical communication will play an even more crucial role, potentially revolutionizing AI chip architecture. Future developments may see “optical computing” replacing traditional electronic computing, enabling more efficient and sustainable AI training and inference.
4. The Future Outlook: Optical Communication + AI = Computing Revolution
-
Photonics + AI Training: Replacing traditional GPUs with photonic chips for faster AI computing.
-
Fully Optical AI Data Centers: AI supercomputing centers based entirely on optical communication to eliminate computational bottlenecks.
-
Optical Chips + Quantum Computing: Exponential growth in AI computing power, paving the way for true AGI (Artificial General Intelligence).
From Being Contained to Full Countermeasures:
A Strategic Move in the Tech War
Five years ago, the U.S. imposed a semiconductor embargo on Huawei to curb China’s technological progress. However, through independent research and development, Huawei gradually overcame these barriers. Now, China’s leadership in optical communication technology has become the Achilles’ heel of U.S. tech dominance.
Unlike the U.S.’s “patchwork” approach to the tech war, China’s actions against NVIDIA are part of a long-term, meticulously planned strategy. If the core of American technology is artificial intelligence, and AI depends on data centers and communication networks, then restricting NVIDIA’s access to advanced optical communication could severely undermine the future of U.S. AI development.
This is one of the most impactful countermeasures in the ongoing tech war.
This move is more than just an antitrust investigation—it is a strategic reconfiguration of global technology. NVIDIA relies on China’s optical communication technology, while China has already taken a leading role in this field. If China were to impose further restrictions on the export of key optical communication components, the U.S. tech industry could face a significant crisis.
News Sources
Reuters: China targets Nvidia with antitrust probe, escalates US chip tensions
The Verge: China opens an antitrust investigation into Nvidia
Barron: Nvidia Stock Drops. Why China Is Probing the AI Chip Maker
Barron: Networking Companies Ride the AI Wave. It Isn’t Just Nvidia.
MIT News: Photonic processor could enable ultrafast AI computations with extreme energy efficiency
Data Centre Magazine: How IBM’s Optical Breakthrough Could Support AI Data Centres
CableLabs: AI and Machine Learning: Lighting the Way for Optical Advancements
Uobam: Fund Focus | china focuses on Al-driven growth
Semianalysis: GB200 Hardware Architecture – Component Supply Chain & BOM
Stephen Xie
Hello, I'm Stephen Xie, a sales representative at Hello Signal. I specialize in Passive Optical Components and Cabinet & Racks. I can recommend the most suitable products at competitive prices, ensuring worry-free service. Feel free to contact me with any questions.
