Did you know? Behind our daily phone scrolling, AI chatting, and watching autonomous driving cars, there are two types of "super chips" - AI computing power chips and photonic chips. They are like the "dual engines" of the digital world, reshaping our future. Today, let's talk about these two technologies that are changing the era.
First, let's talk about AI computing power chips, which are like the "brain engines" of AI. For example, when you chat with ChatGPT, the super-strong computing power behind it relies on AI computing power chips. The most commonly used one currently is the GPU, or graphics processing unit. NVIDIA's products account for over 80% of the global market, almost like the "computing power overlord" in the AI field. Why is it so powerful? Because it is good at processing a large amount of data simultaneously, like having tens of thousands of workers working at the same time, which is extremely fast!
Data shows that the global computing power scale increased from 1397 EFLOPS (which can be simply understood as a "super computing power unit", 1 EFLOPS = 10^18 calculations per second) in 2023 to 16,200 EFLOPS in 2030, an increase of 10 times in 8 years! The Chinese market is also very strong. The intelligent computing power scale is expected to reach 2781.9 EFLOPS in 2028, with an annual growth of nearly 50%. Behind this is the fact that large models (such as GPT-4.5) are becoming smarter and require more computing power for training, as well as scenarios such as cloud computing and autonomous driving that are "consuming" computing power crazily.
In addition to GPUs, there are also ASIC chips (such as Google's TPU), which are like "customized tools" specially optimized for specific tasks, faster and more power-efficient. However, the current U.S. restrictions on high-end chip exports have instead accelerated the rise of Chinese local enterprises, such as Haiguang Information, Cambricon, and DeepSeek Company. They have reduced the training cost of large models a lot, and domestic chips are catching up.
Next is the photonic chip, which is called the "savior of the post-Moore era". Traditional electronic chips rely on electrons to transmit data and now face two major problems: one is the "power consumption wall" - the stronger the computing power, the higher the power consumption. For example, large data centers consume as much electricity as a small city; the other is the "Von Neumann bottleneck" - slow data transmission inside the chip, like a traffic jam on a highway. Photonic chips use "light" instead of "electricity" to transmit and process data using light waves, which is 1,000 times faster than electronic chips and consumes more than 90% less power, equivalent to upgrading from a bicycle to a high-speed train!
Its application scenarios are extremely wide: in data centers, photonic chips are used for "optical interconnection" to solve the problem of data transmission congestion; in AI computing power clusters, they are combined with electronic chips to make computing power fast and power-efficient; in the future, 5G/6G, autonomous driving lidar, and medical detection equipment will all be inseparable from it. It is expected that by 2035, the photonic chip market scale will reach $54 billion, with an annual growth of 44%, which is developing at breakneck speed!
At present, foreign enterprises are leading, but China is also accelerating its breakthroughs. For example, Source Technology has made key progress in 10G EML chips, and Shijia Photon's PLC/AWG chips have the world's first market share. However, we still face challenges such as supply chain restrictions and industry standards and need to continue working hard.
These two types of chips do not fight alone but join forces! AI computing power chips are responsible for super-strong computing, and photonic chips are responsible for ultra-high-speed data transmission and low-power processing, like a "smart brain" with "lightning nerves". For example, in autonomous driving, photonic chips quickly transmit massive data from lidar, and AI computing power chips process and make decisions in real time; in quantum computing, photonic chips can also serve as carriers of "quantum bits" to explore the frontiers of future technology.
In the next 5-10 years, China is expected to build a globally leading chip industry chain. Driven by both policy support and market demand, AI computing power chips and photonic chips will become the key for us to "overtake on a different track". In particular, photonic chips directly bypass the manufacturing bottlenecks of traditional electronic chips, giving us an opportunity to break through Western monopolies.
From AI chatting to autonomous driving, from cloud data centers to future quantum computing, the dual chips are driving a new round of revolution in the digital economy. Are you ready to welcome this intelligent era interwoven with light and electricity?
