Why Is Storage Getting More and More Expensive? What China Truly Lacks Is Not GPUs, But This
发布时间：2026-01-19 10:20:08
Based on the core consensus of the 2025 Data Storage Industry Conference, combined with the latest industry technical reports and localization progress, the optimized promotion copy focuses on "core pain points of AI storage + opportunities for domestic breakthroughs", adding 30 high-value keywords. It not only retains in-depth industry insights but also enhances communication penetration.
The 2025 Data Storage Industry Conference concluded in Guangzhou, gathering Huawei, Sugon, research institutes and leading cloud service providers. This event, which affected the entire industrial chain, finally uncovered two core puzzles: the underlying logic of storage price increases in the AI era, as well as the strategic gaps and golden opportunities in China's storage industry.
First, let’s look at a set of shocking data: in 2025, China's data volume accounted for 30% of the global total, with annual generation exceeding 800EB and continuing to double, and the storage capacity scale reaching 1680EB. While public opinion is fixated on GPUs, the conference reached the most critical consensus: intelligence starts with data, not computing power. Among the three core elements of AI, data has evolved from "information" to a "knowledge base". Computing power determines the speed of operation, but it is storage that serves as the "memory center" determining the upper limit of AI capabilities—without storage, no matter how powerful a GPU is, it is nothing more than an "amnesiac", and no amount of computing is meaningless.
I. The Core Reason for Rising Storage Prices: AI Forces a Revolution in Ultra-High Performance
It’s not that storage is becoming more expensive, but that AI has completely redefined the requirements for storage. Large model training requires injecting PB-level data into GPUs in seconds. Long-text reasoning and multi-round conversations demand storage with microsecond-level latency, 100-million-level IOPS, 100TB/s bandwidth and million-level concurrency. The "100TB bandwidth + 100-million-level IOPS" standard released by Huawei at the conference marks the entry of storage into the "AI-dedicated era"—a completely different level from traditional file storage and backup. Nano-precision technologies such as HAMR (Heat-Assisted Magnetic Recording) and MAMR (Microwave-Assisted Magnetic Recording) have become standard configurations, and the soaring performance threshold has directly driven up costs.
More crucially, there has been a drastic change in data forms: AI has continuously activated cold data that was once dormant, and warm data for high-frequency access has experienced explosive growth. Industry estimates show that warm data has accounted for the core share of total data volume. This type of data, which "cannot be lost and must be accessible at any time", has become the core carrier for AI training and industry applications—yet the underlying media and systems supporting warm data are the only key link that China's storage industry chain has not yet completed.
II. China’s Real Gap in Storage: Strategic Opportunities Amidst 60 Billion Yuan Worth of HDD Dependence
Many people think what China lacks is complete storage devices or main control chips, but the truth is: more than 10 domestic main control chip manufacturers have already mass-produced their products upstream; midstream, enterprise-level storage is led by innovators such as Huawei; downstream, industries including finance, telecommunications and healthcare have already widely adopted domestic storage solutions. The real shortcomings lie in warm storage media and systems, as well as original breakthroughs in underlying basic theories.
As the main carrier for warm data storage, the HDD (Hard Disk Drive) market has exposed the most severe dependency—China's annual HDD consumption reaches nearly 60 billion yuan, yet it is 100% reliant on the oligopoly supply of three foreign companies: Seagate, Western Digital and Toshiba. This monopoly has directly led to a 17% annual price increase for enterprise-level HDDs. Moreover, the explosion of warm data driven by AI has turned this weakness into the "Sword of Damocles" hanging over China's digital economy.
The conference clearly pointed out that warm storage represents the biggest industrial growth point and strategic opportunity in the next decade. Whoever can conquer warm storage media (such as 3D stacked flash memory and optical storage technology) and systems (such as super-node architecture, data fabric and long-memory storage paradigm) will hold the "golden key" to the data field in the AI era. At present, domestic storage has achieved three major breakthroughs: Yangtze Memory Technologies’ QLC chips have reached a single-die capacity of 2Tb; the Chinese Academy of Sciences’ optical storage technology has achieved PB-level single-disk capacity with a 40-year lifespan; the NVMe-oF + RDMA network solution has been widely deployed in core industries, making warm storage the central battlefield for localization efforts.
III. Breaking Through via Industrial Alliance: From Fragmented Efforts to Joint Standard Development
The essence of this conference is an "alliance of the storage industry chain"—faced with the pain points of weak basic theories and "bottlenecks" in warm storage media, the industry has reached a consensus: abandon fragmented competition, join forces to tackle key problems based on common standards, and complete the entire technology chain from media and chips to systems.
Why is this worth paying attention to? Because computing power will depreciate in the future, but data will continue to appreciate in value. For China to become an AI powerhouse, As storage capacity scales doubled from 80 billion yuan in 2021 to 168 billion yuan in 2025, and as Huawei achieved a single-frame bandwidth of 679GB/s in the MLPerf storage benchmark (10 times that of traditional storage), a clear signal has emerged: technological breakthroughs and localization substitution of warm storage are not optional, but imperative.
Key Takeaway: In the next decade, the golden opportunity in the storage field does not lie in following the GPU trend, but in conquering warm storage media and systems! For those who understand, leave a comment with "warm storage" and join us in tracking every step of localization breakthroughs.
Like and follow for more in-depth insights into the AI storage industry!
The 2025 Data Storage Industry Conference concluded in Guangzhou, gathering Huawei, Sugon, research institutes and leading cloud service providers. This event, which affected the entire industrial chain, finally uncovered two core puzzles: the underlying logic of storage price increases in the AI era, as well as the strategic gaps and golden opportunities in China's storage industry.
First, let’s look at a set of shocking data: in 2025, China's data volume accounted for 30% of the global total, with annual generation exceeding 800EB and continuing to double, and the storage capacity scale reaching 1680EB. While public opinion is fixated on GPUs, the conference reached the most critical consensus: intelligence starts with data, not computing power. Among the three core elements of AI, data has evolved from "information" to a "knowledge base". Computing power determines the speed of operation, but it is storage that serves as the "memory center" determining the upper limit of AI capabilities—without storage, no matter how powerful a GPU is, it is nothing more than an "amnesiac", and no amount of computing is meaningless.
It’s not that storage is becoming more expensive, but that AI has completely redefined the requirements for storage. Large model training requires injecting PB-level data into GPUs in seconds. Long-text reasoning and multi-round conversations demand storage with microsecond-level latency, 100-million-level IOPS, 100TB/s bandwidth and million-level concurrency. The "100TB bandwidth + 100-million-level IOPS" standard released by Huawei at the conference marks the entry of storage into the "AI-dedicated era"—a completely different level from traditional file storage and backup. Nano-precision technologies such as HAMR (Heat-Assisted Magnetic Recording) and MAMR (Microwave-Assisted Magnetic Recording) have become standard configurations, and the soaring performance threshold has directly driven up costs.
More crucially, there has been a drastic change in data forms: AI has continuously activated cold data that was once dormant, and warm data for high-frequency access has experienced explosive growth. Industry estimates show that warm data has accounted for the core share of total data volume. This type of data, which "cannot be lost and must be accessible at any time", has become the core carrier for AI training and industry applications—yet the underlying media and systems supporting warm data are the only key link that China's storage industry chain has not yet completed.
II. China’s Real Gap in Storage: Strategic Opportunities Amidst 60 Billion Yuan Worth of HDD Dependence
Many people think what China lacks is complete storage devices or main control chips, but the truth is: more than 10 domestic main control chip manufacturers have already mass-produced their products upstream; midstream, enterprise-level storage is led by innovators such as Huawei; downstream, industries including finance, telecommunications and healthcare have already widely adopted domestic storage solutions. The real shortcomings lie in warm storage media and systems, as well as original breakthroughs in underlying basic theories.
As the main carrier for warm data storage, the HDD (Hard Disk Drive) market has exposed the most severe dependency—China's annual HDD consumption reaches nearly 60 billion yuan, yet it is 100% reliant on the oligopoly supply of three foreign companies: Seagate, Western Digital and Toshiba. This monopoly has directly led to a 17% annual price increase for enterprise-level HDDs. Moreover, the explosion of warm data driven by AI has turned this weakness into the "Sword of Damocles" hanging over China's digital economy.
The conference clearly pointed out that warm storage represents the biggest industrial growth point and strategic opportunity in the next decade. Whoever can conquer warm storage media (such as 3D stacked flash memory and optical storage technology) and systems (such as super-node architecture, data fabric and long-memory storage paradigm) will hold the "golden key" to the data field in the AI era. At present, domestic storage has achieved three major breakthroughs: Yangtze Memory Technologies’ QLC chips have reached a single-die capacity of 2Tb; the Chinese Academy of Sciences’ optical storage technology has achieved PB-level single-disk capacity with a 40-year lifespan; the NVMe-oF + RDMA network solution has been widely deployed in core industries, making warm storage the central battlefield for localization efforts.
III. Breaking Through via Industrial Alliance: From Fragmented Efforts to Joint Standard Development
The essence of this conference is an "alliance of the storage industry chain"—faced with the pain points of weak basic theories and "bottlenecks" in warm storage media, the industry has reached a consensus: abandon fragmented competition, join forces to tackle key problems based on common standards, and complete the entire technology chain from media and chips to systems.
Why is this worth paying attention to? Because computing power will depreciate in the future, but data will continue to appreciate in value. For China to become an AI powerhouse, As storage capacity scales doubled from 80 billion yuan in 2021 to 168 billion yuan in 2025, and as Huawei achieved a single-frame bandwidth of 679GB/s in the MLPerf storage benchmark (10 times that of traditional storage), a clear signal has emerged: technological breakthroughs and localization substitution of warm storage are not optional, but imperative.
Key Takeaway: In the next decade, the golden opportunity in the storage field does not lie in following the GPU trend, but in conquering warm storage media and systems! For those who understand, leave a comment with "warm storage" and join us in tracking every step of localization breakthroughs.
Like and follow for more in-depth insights into the AI storage industry!
