Tech giants like Google, Meta, and Amazon are aggressively advancing ASIC chips, which is not a technological revolution but an inevitable result of the restructuring of the global macro landscape. In my view, this is akin to the Three Kingdoms, where great powers vie for dominance while smaller nations suffer.
ASIC chips have become a battleground for global technological hegemony. Google’s TPU shipment has been raised from 1.8 million to 2.7 million units, while Meta has progressed from 5nm to 3nm and then to 2nm processes, with an average selling price increasing fivefold; these are not coincidences.
The reason these giants are so fervently investing in ASICs is that they understand that the era of general-purpose computing is coming to an end. As the parameter scale of AI large models surpasses trillions, traditional GPU architectures are encountering bottlenecks in terms of energy efficiency and memory bandwidth.
Google’s TPU has enhanced matrix operation efficiency to 30 times that of GPUs through pulse array design; Cambricon’s Siyuan 590 chip employs sparse computing technology, reducing the training cost of large models by 30%. This is why these giants are willing to invest heavily in self-developed chips.
The Fantasy and Reality of Domestic Substitution
Many people are blindly optimistic that Chinese ASIC chips can quickly achieve domestic substitution, but caution is warranted. The macro environment has changed, and de-globalization is a major trend.
Although the technological level of Chinese ASIC companies is continuously improving, with products like Huawei’s Ascend and Cambricon reaching internationally advanced levels in AI computing power and energy efficiency, we still face risks of technological blockade in high-end processes, EDA tools, and IP cores.
Data shows that under the backdrop of U.S. export controls, the market share of Chinese ASIC companies is expected to rise from less than 20% in 2024 to 40% in 2025. However, this is mainly driven by policy and internal market substitution, rather than true global competitiveness.
We cannot seek a sword by carving a boat. The present will not repeat the past, as the macro environment has completely changed. The debt ratio of Chinese residents is nearing that of the U.S. before the subprime mortgage crisis. This means we have little room for maneuver internally.
Investment Landscape of Coexisting Risks and Opportunities
Do not bet on a single direction, but rather diversify across multiple scenarios to manage risk. The same mindset is needed in the ASIC investment field.
The opportunities for domestic ASICs lie in
- Strong policy support, as the Ministry of Industry and Information Technology has released the “Action Plan for the Development of the Integrated Circuit Industry,” clearly stating the goal of achieving over 60% domestic production rate of ASIC chips in AI training and intelligent driving by 2028.
- Explosive market demand, where ASIC chips are expected to become the mainstream choice in data centers and edge computing scenarios by 2025 in China.
- Technological breakthroughs, with new architectures like storage-computing integration and photonic computing expected to account for over 30%, achieving energy efficiency multiple times that of traditional chips; 3D integration technology will enable high-level stacking, significantly increasing storage bandwidth.
However, risks cannot be ignored
- Risks of technological blockade, as we still face technological blockades in high-end processes, EDA tools, and IP cores.
- Speed of iteration disparity, as foreign giants typically iterate once a year, while domestic products like Ascend 910B and Cambricon 590 are slower to update.
- Lagging ecosystem development, as the CUDA ecosystem remains monopolistic, and domestic chips lack comprehensive software ecosystem support.
Strategic Thinking
Long on essential consumption in China, short on discretionary consumption. Due to monetary easing leading to increased wealth disparity, most people have no money and can only purchase essential goods, lacking funds to buy cars, phones, or home appliances.
In the ASIC field, I adopt a relative value strategy, going long on ASIC companies with strong domestic substitution capabilities, such as Xingsen Technology, and shorting companies that rely entirely on imported technology; going long on companies with independent IP and ecosystems, and shorting those that solely depend on process advancements.
Focus on three types of companies: first, high-end ASIC design firms capable of developing high-end products like AI training chips and vehicle domain controllers; second, key IP core suppliers that form technological barriers in NPU, ISP, and interface protocols; third, Chiplet packaging and testing companies that master core technologies in 2.5D/3D packaging.
The Future Outlook is a Brutal Survival Game
In the next five years, ASIC chips will deeply integrate into the bloodstream of the digital economy, becoming the core engine driving industrial upgrades. However, this path will not be smooth.
By 2030, the Chinese ASIC chip market is expected to maintain an annual growth rate of 25%, exceeding 1.2 trillion yuan. However, this is merely the market scale and does not represent profit scale, nor does it indicate international competitiveness.
We need to be clear that the China-U.S. trade conflict is a long-term battle that requires 10-20 years of preparation. In the ASIC field, this conflict is even more direct and brutal.
Do not blindly worship technological myths, and do not be overly optimistic about domestic substitution.
The current predictions of domestic substitution for ASICs are a possibility, but not a highly probable event.
In this uncertain era, we need to maintain independent thinking and reject blind worship and myths. Only in this way can we find real investment opportunities in this ASIC chip war, rather than becoming the chives harvested by others.