The latest earnings report from NVIDIA, a global AI bellwether, has once again sent ripples through capital markets. Despite impressive revenue and guidance, its stock initially rose over 5% on the earnings day but closed down 3.15%. By Friday, NVIDIA had declined 0.97%, bringing its November losses to nearly 12%.
This volatility reflects market concerns about the sustainability of its future growth, compounded by external uncertainties. The pressure on U.S. tech stocks has also impacted the A-share market. On November 21, the computing power sector fell 3.38%, with net outflows of 14.12 billion yuan. Key players like New H3C, TFC, and InnoLight dropped 8.46%, 7.34%, and 5.69%, respectively. The AI chip sector declined 4.33%, with Cambricon down 5.54%.
Fund holdings in electronics and communications sectors (including computing power-related stocks) reached historical highs by Q3 2023, creating complex positioning dynamics. Analysts note that year-end liquidity needs may be driving profit-taking.
Experts suggest the market is reassessing valuations relative to fundamentals in the computing power sector. However, long-term trends remain intact, with tech giants like Microsoft, Google, Meta, and Amazon continuing to increase capital expenditures, driving demand for optical modules and servers.
In China, the computing power industry is undergoing transformation across the value chain, from chip manufacturing to system integration, with a focus on self-sufficiency. Domestic innovation is accelerating, supported by strong AI computing demand. While overseas markets lead growth, Chinese internet companies are adjusting investments based on GPU efficiency and utilization.
Local chip suppliers are gaining ground, with Cambricon reporting revenue of 4.607 billion yuan in the first three quarters, up 2,386.38% year-on-year, and turning a profit of 1.605 billion yuan. Hygon also saw significant growth, with revenue up 54.65% to 9.49 billion yuan. These improvements reflect technological breakthroughs and commercialization progress.
Policy support is bolstering domestic computing power development, with national initiatives and procurement preferences creating a favorable environment. The industry is also shifting from hardware expansion to efficiency optimization. Research shows that AI model capability density doubles every 3.5 months, reducing inference costs dramatically. Algorithm optimizations are improving hardware utilization, with some methods cutting costs by 56% while boosting accuracy.
System-level innovations, like hyper-node technology, are helping bridge performance gaps in domestic GPUs. Software ecosystem development is also advancing, with tailored algorithms enhancing hardware performance. Cooling technologies are evolving to meet higher power demands, with liquid cooling expected to grow at a 46.8% CAGR in China through 2029.
Investment opportunities are emerging across the computing power value chain. High-speed optical modules (800G/1.6T), optical chips, memory, and PCBs are key focus areas. Domestic AI chips from Huawei, Cambricon, and Hygon are accelerating iteration and adoption. The industry is maturing, with investment shifting from infrastructure to application innovation.
The computing power sector is transitioning from scale-driven growth to efficiency and self-sufficiency. Companies that adapt to these trends will likely thrive in the next phase of industry evolution.
Comments