<img src="https://imageproxy.pbkrs.com/https://inews.gtimg.com/om_bt/Oywm_vw3REjokgX8jN5sux5bYzWibrxAvnuhXpBh-xIb4AA/641?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://inews.gtimg.com/om_bt/Oywm_vw3REjokgX8jN5sux5bYzWibrxAvnuhXpBh-xIb4AA/641"/>
OpenClaw has been continuously popular recently. As an open-source AI agent framework, it grants large models local system operation permissions, enabling them to manage schedules, read and write files, execute scripts, and complete complex task loops, thus becoming an all-weather &#34;digital employee&#34; and realizing the &#34;autonomous execution&#34; mode of AI applications.
Boosted by this, on March 10, the AI industry chain collectively strengthened, with AI computing power sectors such as optical modules CPO, optical communication, and PCB showing strong rebounds. As of the time of writing, the largest communication ETF by index scale, Huaxia (515050), surged again in the afternoon, rising over 4.6% during the session. Among the holdings, Yuanjie Technology (688498.SH) and Tianfu Communication (300394.SZ) rose over 10%, while Guangxun Technology (002281.SZ), Dongshan Precision (002384.SZ), and Fenghuo Communication (600498.SH) saw multiple stocks hitting the daily limit. The low-fee ChiNext AI ETF Huaxia (159381) rose over 2.5%, and the Cloud Computing ETF Huaxia (516630) aimed for a third consecutive increase.
OpenClaw Drives Explosive Growth in Computing Power Demand
The sustained tight demand for computing power is primarily driven by the intensive upgrades of large models both domestically and internationally, especially the deep evolution of agents and multimodal applications, which has led to an exponential increase in inference computing power consumption, becoming a new growth point for computing power demand.
Similar to the emergence of phenomenal products like Kimi and DS, OpenClaw brings full industry chain investment opportunities from infrastructure to downstream applications to the capital market. Currently, the overall AI application market is relatively sluggish, but the popularity of OpenClaw has certain expectations that may reverse the market's pessimistic outlook.
The computing power infrastructure has a high degree of certainty. The operational model of OpenClaw leads to a leap in computing power consumption levels. Unlike traditional Q&amp;A-style AI that consumes hundreds of tokens, its execution of complex tasks requires the model to continuously think, plan, and call tools, with a single task potentially consuming tens of thousands or even millions of tokens. The demand for computing power shifts from intermittent &#34;dialogue&#34; modes to continuous &#34;execution&#34; modes, driving explosive growth in inference-side computing power demand, benefiting cloud computing service providers, AI server manufacturers, and domestic computing power chip manufacturers.
Trend of Rising Computing Power Prices Emerges
As early as before, large model service providers had already raised prices first, directly reflecting the overload of computing power demand. On February 12, Zhipu (02513.HK) officially released a price adjustment notice for the GLM Coding Plan, clearly canceling the first purchase discount and raising package prices overall by 30% or more. Behind the official statement of &#34;increased market demand, user scale, and call volume&#34; is the passive price adjustment situation due to the continuous tightening of computing power resources and rising operational costs of models. This action also sends a clear signal to the market: the computing power demand of leading large model enterprises has already reached its peak and is still climbing rapidly. This wave of demand is transmitting upward along the industry chain, layer by layer, to every core link of upstream computing power hardware
Moreover, the prices of core hardware components have skyrocketed, and the capacity gap continues to widen. Driven by the explosive demand in the AI server market, the prices of DRAM and NAND memory have continued to rise, with the price increase for storage used in consumer products exceeding 600% over the past year, completely reversing the previous downward cycle and becoming the core driving force behind the recovery of the storage industry, highlighting the explosive demand in the storage segment. At the same time, NVIDIA (NVDA.US) high-end GPU chips have long been in a state of &#34;hard to find,&#34; with some flagship models seeing price increases of over 50% in the secondary market, and the official delivery cycle lasting as long as 3-6 months. Even as the new generation of chips gradually begins to ship, it will be difficult to alleviate the global supply tightness in the short term, and the supply-demand imbalance in computing hardware will continue.
In terms of segmentation, the supply-demand pattern in the optical fiber and cable industry has reversed. After experiencing a two-year destocking cycle, the optical fiber and cable industry has entered a new round of prosperity. By the end of 2025, the capacity utilization rates of the four leading companies in the industry will remain high. Among them, the restructuring of demand is the key to driving up optical fiber prices. The surge in demand for high-end optical fibers such as G657A1 and G657A2 in AI infrastructure and drone fields has led to more capacity being tilted towards high-end products, resulting in a supply shortage of the most basic G652D optical fibers. Currently, the average price of G652D optical fibers has risen for two consecutive quarters, with the market price for bare fibers reaching a peak of 40 yuan per core kilometer in January 2026. According to statistics, global optical fiber demand is expected to climb to 880 million core kilometers by 2027, with the proportion of data center internal and DCI scenarios rising from less than 5% in 2024 to 35%. As overseas manufacturers' capacities increasingly favor the North American market, global optical fiber supply will continue to tighten, and leading domestic companies with export capabilities are expected to achieve significant performance profit elasticity.
Under the wave of computing power expansion, these segments will continue to benefit.
The pattern of high global demand for computing power is unlikely to change in the short term. With the shipment of NVIDIA's new generation of chips, the continuous upgrade of domestic and foreign large models, and the accelerated construction of computing power infrastructure, multiple segments of the computing power industry chain will continue to benefit, specifically focusing on the following directions:
<ol>
<li>
AI Chips (Core Driving Force): As the &#34;heart&#34; of computing power, NVIDIA's GPUs remain absolutely core. The shipment of its Blackwell series chips will directly drive its own performance growth, while domestic AI chip companies will also benefit from the adaptation needs of domestic large models, gradually achieving import substitution and alleviating the tight supply of computing power domestically.
</li>
<li>
Optical Modules (Core of Computing Power Transmission): As the scale of computing clusters expands, the demand for data transmission within and across data centers has surged. Optical modules, as core transmission devices, especially CPO (Co-Packaged Optics) technology, can significantly improve transmission efficiency and reduce power consumption, adapting to the needs of high-end computing clusters.
</li>
<li>
Storage Devices (Data Carriers for Computing Power): Large model training and inference require massive data storage. The continuous rise in prices of DRAM and NAND memory confirms the strong demand in the storage segment. As computing power expansion continues, the demand for storage devices will remain high
</li>
<li>
Server and Computing Power Infrastructure: AI servers are the core carriers of computing power implementation. As cloud vendors and large model companies continue to increase their investments in computing power, the demand for AI servers will continue to explode. At the same time, related aspects such as data center construction and computing power cluster building will also benefit from the global expansion of computing power.
</li>
</ol>
For ordinary investors, the computing power sector has many sub-segments with significant individual stock volatility. Using ETFs for investment has become an efficient and convenient choice, as it can comprehensively cover the core targets of the computing power industry chain while diversifying individual stock risks. Relevant computing power-related ETFs to pay attention to include:
ChiNext AI ETF - Huaxia (159381): The index tracked has half of its weight concentrated in the optical module CPO sector, while the other half covers the AI software application field, forming a balanced layout of &#34;hardware + application.&#34; The top 10 weighted stocks are Zhongji Xuchuang (300308.SZ) (13.23%), Xinyi Sheng (300502.SZ) (12.67%), Tianfu Communication (10.18%), Runze Technology (300442.SZ), BlueFocus Communication Group (300058.SZ), Xiechuang Data (300857.SZ), Kunlun Wanwei (300418.SZ), Beijing Junzheng (300223.SZ), Wangsu Science &amp; Technology (300017.SZ), Runhe Software (300339.SZ). The current fund size is nearly 2 billion yuan, with an on-market comprehensive fee rate of only 0.20%, the lowest in its category, suitable for investors seeking high elasticity and optimistic about the AI + main line. Off-market connection (Class A: 025505; Class C: 025506).
Communication ETF - Huaxia (515050): Tracks the CSI 5G Communication Theme Index, with the latest size nearing 8 billion yuan, deeply covering the optical module, optical fiber and cable, communication equipment, and memory sectors of computing power infrastructure. The combined weight of CPO + CPB concept stocks exceeds 76%, ranking first in the entire market. The top 10 weighted stocks include Zhongji Xuchuang, Xinyi Sheng, Tianfu Communication, Dongshan Precision, Hengtong Optic-Electric (600487.SH), as well as covering Luxshare Precision (002475.SZ), Industrial Fulian (601138.SH), Zhaoyi Innovation (603986.SH), balancing growth and stability, suitable for investors looking to make long-term investments in computing power infrastructure. (Class A: 008086; Class C: 008087).
Cloud Computing ETF - Huaxia (516630): Focuses on domestic AI software and hardware computing power, with a total weight of computer software + cloud services + computer equipment reaching 83.7%. DeepSeek and AI application content both exceed 40%, accurately covering the core segments of the domestic computing power industry chain, benefiting from the trend of domestic large model upgrades and self-controllable computing power. The on-market comprehensive fee rate is only 0.20%, the lowest in its category, suitable for investors optimistic about the development of domestic computing power. Off-market connection (Class A: 019868; Class C: 019869)
Note: The ETFs mentioned above do not charge subscription fees, redemption fees, or sales service fees. Subscription and redemption agents may charge a commission not exceeding 0.5%, which includes related fees charged by securities exchanges, registration and settlement institutions, etc.; in addition, the management fee rates and custody fee rates for the relevant ETFs and their linked products are as follows:
<img src="https://imageproxy.pbkrs.com/https://inews.gtimg.com/om_bt/OKqXtcPONYzYcmaI8s9TCbW5ItXcdZW7EeiWE-QgMMxjYAA/641?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://inews.gtimg.com/om_bt/OKqXtcPONYzYcmaI8s9TCbW5ItXcdZW7EeiWE-QgMMxjYAA/641"/>

688047

688158

DTCR

OpenClaw, as an open-source AI intelligent framework, has driven explosive growth in the demand for inference-side computing power. It grants large models local system operation permissions, enabling them to perform complex tasks and become all-weather "digital employees." As a result, the AI industry chain has collectively strengthened, with related ETFs and individual stocks performing outstandingly. The surge in computing power demand mainly stems from the upgrades of large models both domestically and internationally, particularly the evolution of intelligent agents and multimodal applications, which is expected to benefit cloud service providers and domestic computing power chip manufacturers

腾讯新闻 - 财经

LOONGSON TECHNOLOGY

UCloud

Global X Data Center & Dgtl Infrs ETF

GTJA Allianz SSE STAR Chip Design Thematic ETF

ROBO Global Robotics

Roundhill Generative AI & Tech ETF

Global X Funds Global X Artificial Intelligence & Tech ETF

Guotai SSE STAR Chip ETF

Invesco Semiconductors ETF

Direxion Semicon Bull 3X

ISHRS S&P Glb It

VanEck Semiconductor ETF

iShares Semiconductor ETF

E Fund CSI Cloud Computing and Big Data Theme ETF

Guotai CES Semiconductor Chip Industry ETF

KraneShares 2x Long BABA Daily ETF

Global X Robotics & Artificial

BABA 2x Long Daily ETF - GraniteShares

SPDR S&P Software

Franklin Exponential Data ETF

GX CN CLOUD

SPDR S&P Semicon

Harvest SSE STAR Chip ETF

iShares Future AI & Tech ETF

China Southern CSI Semiconductor Industry Custom ETF

ChinaAMC Guozheng Semiconductor Chip ETF

Guotai CSI Semiconductor Material Equipment Theme ETF

FT Nasdq AI & Robotics

Proshares Big Data Refiners ETF

Tianhong CSI Computer Theme ETF

Guotai CSI Computer Theme ETF

ChinaAMC SSE STAR Semiconductor Material & Equipment Thematic ETF

GTJA Allianz CSI All Share Semiconductors & Semiconductor Equipment ETF

iShares Expanded Tech Software Sector ETF

Global X Cloud Computing ETF

BABA-W

Alibaba

OpenClaw is booming, driving explosive growth in demand for inference-side computing power