英伟达的生死线，根本不是芯片：卡死全球AI算力的4大材料命脉

Core Insights - The competition in AI computing power has shifted from chip design to the underlying materials, which are critical for performance and cost efficiency [3][6][66] - The article identifies four key material sectors that are crucial for AI computing: advanced packaging materials, cooling materials, optical interconnect materials, and new storage materials [11][63] Advanced Packaging Materials - The core barrier in advanced packaging is not the technology but the material system, with materials accounting for 65% of costs, particularly ABF substrates [14][15] - Domestic manufacturers face significant challenges due to strict quality control, patent barriers, and a tightly bound supply chain dominated by foreign firms [16][17][19] - Current domestic production rates for high-precision ABF substrates are below 3%, indicating a long way to go for self-sufficiency [20] Cooling Materials - Cooling materials are essential, constituting over 30% of the total cost of AI data centers, with thermal management materials making up more than 60% of that cost [23][24] - The market for immersion cooling liquids is heavily dominated by foreign companies, with domestic production rates for high-end products below 10% [29] - The lack of compatibility verification and high costs associated with testing new materials hinder domestic manufacturers from entering the supply chain [30] Optical Interconnect Materials - The demand for optical interconnects is driven by the exponential growth in bandwidth requirements for AI models, with silicon photonics being the ultimate solution [34] - Domestic manufacturers hold over 60% of the global market for high-speed optical modules, but the localization rate for core optical materials is below 30% [40] - The precision required for manufacturing optical fibers and waveguide materials presents significant barriers for domestic firms [36][38] New Storage Materials - The "memory wall" issue in AI training necessitates breakthroughs in new storage technologies, with a focus on integrated storage and computing solutions [45] - Current domestic commercialization rates for new storage materials are below 5%, with most developments still in the laboratory stage [51] - The reliance on foreign patents and the inability to achieve consistent production quality are major hurdles for domestic manufacturers [47][49] Industry Challenges and Opportunities - The article emphasizes that the primary challenge for China's new materials industry is not technological capability but the lack of a complete industrial ecosystem [57] - A collaborative approach involving policy support, leading companies, and coordinated supply chains is essential for breaking the cycle of dependency on foreign materials [61][62] - Investors are encouraged to focus on companies that have achieved material breakthroughs and are entering supply chains, rather than those that merely present concepts without proven capabilities [67]