AI Power Revolution: Supercapacitors Emerge as Next Price Surge Category; LIC Technology Gains Edge as Mainstream Choice

Stock News
May 21

Sinolink Securities Co., Ltd. released a research report stating that the leap in computing density within AI data centers is driving a transformation in power supply architecture, positioning supercapacitors as essential structural components. AI workloads are shifting from steady-state to millisecond-level pulse transitions, placing traditional three-tier backup power systems under comprehensive pressure regarding response speed, cycle life, and energy loss. Supercapacitors achieve microsecond-level response and ultra-long cycle life through mechanisms like electric double-layer physical adsorption or lithium-ion intercalation. Among these, the Lithium-Ion Capacitor (LIC) route, with its higher energy density and compact form factor, is better suited to the spatial constraints of AI server racks and is becoming the mainstream choice. The main viewpoints of Sinolink Securities Co., Ltd. are as follows:

Starting from the GB300, NVIDIA has integrated electrolytic capacitors into the power shelf, which can reduce peak grid demand by approximately 30%. The next-generation Rubin platform will significantly increase energy storage capacity compared to its predecessor, marking the upgrade of energy storage components from auxiliary functions to core system elements. Domestic manufacturers like Jianghai Capacitor have confirmed that their MLPC and supercapacitor products are compatible with the GB300 solution and are accelerating integration into the server supply chain. Supercapacitors, lithium batteries, and diesel generators form a multi-level complementary system characterized by "fast but short duration—slow but long duration—long but slow response," with all three being indispensable.

The report suggests that supercapacitors have transitioned from laboratory solutions to standard rack-level configurations. The volume ramp-up of the Rubin platform in the second half of 2026 will be a critical window for performance realization across the related industrial chain. The LIC and EDLC (Electric Double-Layer Capacitor) routes are developing in parallel, with Musashi and Maxwell leading in their respective fields. Supercapacitors are currently primarily divided into two technological routes: EDLC, which relies purely on physical electric double-layer energy storage and offers advantages in long lifespan, high-rate capability, and strong pulse response; and LIC, which incorporates a lithium-ion intercalation mechanism, providing higher energy density and smaller volume, making it more suitable for high-density AI server scenarios. The NVIDIA GB200/GB300 AI server supply chain is accelerating the adoption of the LIC/HSC (Hybrid Supercapacitor) route. Japan's Musashi, with its HSC product line, has become a core supplier, while Maxwell has long focused on the EDLC route, holding a leading advantage in high-power transient response applications. Overall, LIC and EDLC are expected to coexist long-term in applications such as AI servers, energy storage, and grid frequency regulation, with substitution relationships forming in certain use cases.

The GB300 is propelling supercapacitors into an era of standard configuration, creating a window of opportunity for domestic manufacturers to fill gaps. With the increased power of the single GB300 NVL72 rack, AI servers' requirements for transient power supply stability have risen significantly. Supercapacitors and BBUs (Battery Backup Units) are being upgraded from optional components in the GB200 era to standard parts of the power architecture in the GB300 era, and are being integrated into the unified Energy Storage Tray system. The current mainstream solution in the supply chain is dominated by the CESS system, a collaboration between Musashi and Flex. However, against the backdrop of exploding AI server demand, a significant supply gap has emerged. The demand for supercapacitors corresponding to the GB300 is substantial, and current production capacity may be insufficient. Assuming GB300 NVL72 rack shipments reach 50,000 to 60,000 units in 2026, with each GB300 rack requiring 5 BBU modules and over 300 supercapacitors, the estimated demand for supercapacitors for the GB300 in 2026 would be 15 to 18 million units. In contrast, Musashi's planned annual capacity by Q3 2026 is 6.5 million units. This presents a significant opportunity for domestic manufacturers to fill the gap. Companies including Dongyangguang, Jianghai Capacitor, and Siyuan Electric (through its subsidiary Xijing Carbon Energy) are actively developing in the EDLC/hybrid supercapacitor direction and are poised to benefit from the industrial opportunities brought by the AI power architecture upgrade.

Related investment targets include: Supercapacitors: Dongyangguang, Jianghai Capacitor, Siyuan Electric, Haixing Shares, and Aihua Group. SST (Solid-State Transformer): Sifang Co., Ltd., Jinpan International Group, Sungrow Power Supply, Jingquanhua, and Click Technology. SiC (Silicon Carbide) required for SST: Tianyue Advanced, Jingsheng Co., Ltd., Yujing Co., Ltd., and Sanan Optoelectronics.

Risk warnings include: intensified industry competition; slower-than-expected progress in technology research and development; and risks associated with cyclical fluctuations in capital expenditures within specific downstream industries.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10