Three Major Technology Trends: Chiplet, RISC-V, AI Co-Design

When discussing future technology, the market is most easily captivated by grand narratives: Artificial General Intelligence (AGI), quantum computing, fully autonomous robots—all sound exciting. However, what will truly achieve large-scale implementation in the next five years and reshape industry structures are often not the flashiest concepts, but rather technologies that have crossed the laboratory threshold and are entering a phase of accelerated commercialization. If we focus back on the semiconductor and computing industries, Chiplet, RISC-V, and AI Co-Design are likely to be the three most noteworthy trends.

First is Chiplet. In the past, chip design emphasized integrating all functionalities into a single large chip. But as advanced process costs soar and yield challenges intensify, the traditional single-chip approach is beginning to hit economic limits. The core value of Chiplet isn’t just about 'making chips smaller,' but modularizing different functions so that elements like CPU, GPU, I/O, and memory controllers can be combined more flexibly. This approach offers two immediate benefits: one, it reduces R&D and manufacturing risks; two, it accelerates product iteration. For large chip companies, Chiplet allows high-performance computing products to reach the market faster; for mid-sized design firms, it lowers the barrier to entry into the high-end market. Over the next five years, servers, AI accelerators, automotive chips, and even high-end edge devices may widely adopt this 'building block' design. The focal point of semiconductor competition will shift from single-node processes to packaging, interconnects, and system integration capabilities.

The second trend is RISC-V. Its growing attention is not solely due to its streamlined technical architecture, but also because it offers an openness rarely seen in the industry. For a long time, the processor core market has been highly concentrated, with companies developing their own chips often constrained by licensing fees, development restrictions, and geopolitical risks. As an open instruction set architecture, RISC-V allows enterprises to tailor designs based on their specific needs—from IoT, industrial control, wearables, to automotive systems and AI edge computing—providing greater autonomy. This doesn't mean it will completely replace existing architectures in the short term, but it is highly likely to expand rapidly in 'specific applications and scenarios.' Especially as countries increasingly emphasize supply chain security and technological independence, RISC-V's business appeal is no longer just an engineering issue but a strategic industry question. Over the next five years, its most probable development path won't be full disruption but steady penetration—first solidifying itself in embedded and specialized chips before advancing towards higher-end applications.

As for AI Co-Design, it represents another profound transformation: future competition won’t just be about 'how fast chips run,' but whether hardware, models, and software can be co-designed and mutually optimized. In the past, hardware design and software development often operated at different paces, with general-purpose hardware being developed first and software adapting afterward. But in the AI era, model updates happen too quickly and computational demands are too high for this division to sustain efficiency. Consequently, more and more companies are starting to reverse-engineer hardware architectures based on model requirements, optimizing compilers, memory configurations, and data flow designs simultaneously. This is the essence of AI Co-Design: treating the system as a whole rather than piecing together components. The rapid maturation of this approach stems from the fact that generative AI has directly turned computational power costs into enterprise costs. Whoever can deploy models with lower power consumption and latency gains an advantage in cloud services, smart terminals, and enterprise AI markets.

Notably, these three trends do not exist in isolation but instead reinforce each other. Chiplet provides more flexible hardware integration, RISC-V offers greater architectural autonomy, and AI Co-Design pulls application requirements directly into the chip design process. Where these three intersect lies the heart of the next wave of semiconductor innovation: more modularity, more openness, and closer alignment with real-world scenarios. The winners over the next five years may not necessarily be companies with a single cutting-edge technology, but those capable of combining these three capabilities into viable business solutions.

(Chip and Computing Power Series No. 41)

$NVIDIA (NVDA.US)$ $Taiwan Semiconductor (TSM.US)$ $Amazon (AMZN.US)$ $Alphabet-C (GOOG.US)$ $Alibaba (BABA.US)$

251K Views