Over the past two years, the flash memory market has encountered a challenging period characterized by stagnationHowever, a transformation began in the last quarter of 2023, driven by an increase in the availability of flash memory resources, which spurred a significant rise in the prices of flash productsThis shift has prompted discussions about whether the Chinese flash memory market will fully recover in 2024 or experience a drastic downturn once againThe ongoing AI boom in China adds another layer of complexity, raising questions about its potential effects on the flash market, such as the anticipated mainstream adoption of QLC SSDs this yearFurthermore, technological advancements such as PCI-E 5.0 and CXL related to flash memory are anticipated to develop new trends in the industry as well.

Recently, a summit focusing on the flash memory market was held in China, where representatives from numerous global flash memory enterprises shared insights into current market dynamics and technological advancements

These discussions provided a valuable perspective on how the market is evolving, especially with the increasing relevance of AI.

As the trend of large AI models continues to gain momentum, the demand for computational power has taken center stageYet as the AI landscape evolves, numerous challenges related to data storage have surfacedDuring the summit, key players in the industry such as Solidigm, SK Hynix, and Samsung addressed the implications of AI on storage systems, highlighting the multifaceted impact AI has on storage requirements.

The influence of AI on storage is expected to be both significant and far-reachingFor starters, the rise of multimodal AI models brings about an explosion of data, resulting in exponential growthIn fact, Gartner predicts that by 2025, AI-generated data will account for 10% of all data generated globally

• Samsung to Integrate SK Hynix Technologies
• The Impact of Interest Rate Decisions on the Bond Market
• Economic Drivers of Global Stock Markets
• Decoding Economic Indicators
• Europe's Economic Influence on FX Markets

Furthermore, storage systems will have to meet performance demands tied to GPUs, thus enhancing GPU utilization effectivelyAdditionally, with ongoing improvements in storage capacity density, challenges associated with power consumption, physical space, and costs cannot be overlooked.

According to Solidigm's Vice President of Sales for the Asia-Pacific region, Ni Jinfeng, certain reports have shown that in some specific scenarios, storage power consumption can be as much as 35% of the total server power consumptionAs the AI landscape progresses rapidly, addressing power consumption associated with storage is inevitably becoming a pressing challenge.

The evolution of large-scale AI models signifies a profound shift towards managing massive quantities of high-quality data and ensuring efficient data processing, which creates a ripple effect across various environments such as data centers and edge computing

This is further accentuated by the increasing competitive landscape of AI development within China, known as the “Hundred Model War”. It becomes evident that the influence of AI on storage will only intensify from this point forward.

On analyzing the demands brought forth by AI, one notices radical changes in storage requirements, eventuating necessities for improved throughput, capacity, and scalabilitySo what strategies can be adopted to meet these evolving data storage demands?

Currently, there are two main technological pathways: one, involving high-capacity hard drive technologies that focus on enhancing storage capacity through increased platter density and additional platters; and the second, utilizing flash memory, which offers high capacity and low power consumption, achieving a balance among costs, power usage, physical space, and capacity.

Ni Jinfeng assesses that the rapid advancement of AI technology is revolutionary for storage needs, asserting that the advantages of flash memory will increasingly emerge in terms of performance, capacity, and total cost.

From a market perspective, the expansive investments in AI technologies in North America have sparked a considerable demand for ultra-high-capacity SSDs

Ni forecasts that the demand for high-capacity SSDs within the domestic market will catch up swiftly in 2024.

The recovery of the storage market is directly reflected in the increasing capacity requirementsWithin the realm of flash memory, capacity density has established a robust foundation in 2023. Companies like Solidigm, SK Hynix, Yangtze Memory Technologies, and Samsung have advanced the layers of NAND to unprecedented heightsFor instance, Solidigm has aggressively promoted QLC SSD products in the market, launching a staggering 61.44TB QLC SSD.

Industry experts anticipate that QLC SSD will see significant validation across various parameters including quality, reliability, and lifespan, establishing itself as the mainstream choice in 2024.

In terms of practical application, QLC SSDs are rapidly penetrating consumer and mobile markets

The emergence of AI concepts within smartphones and PCs, spurred by advancements in large model technology, has a high likelihood of accelerating the deployment of QLC SSD productsDue to the need for capacity reduction in production, the QLC SSD market may face a supply-demand imbalance in the current year; nevertheless, in the long run, sustained demand will undoubtedly propel the QLC SSD market forward in the absence of production increase.

Moreover, in the enterprise market, the current proportion of SSD utilization in Chinese data centers remains lowThe advancements in technologies such as ZNS, PCI-E, and CXL, coupled with the stimulation driven by applications in AI and big data, signify that QLC SSDs are poised to become mainstreamNumerous storage array manufacturers, including DELL EMC, NetApp, and Inspur, have launched QLC SSD-based storage products in 2023. As these manufacturers fully embrace QLC SSD technology, it is anticipated that QLC SSD-based storage solutions will make their way into more enterprise data centers by 2024.

Additionally, the combination of PCI-E 4.0 and QLC SSDs has already seen substantial application in data center environments involving read-intensive and certain read-write mixed workloads

With the commercialization of PCI-E 5.0 products on the horizon, the momentum surrounding the adoption of QLC SSDs is expected to build further.

Ni believes that both consumer and enterprise-level markets are adequately prepared for this transitionIn the consumer landscape, QLC SSDs are set to accelerate the replacement of TLC, while in the enterprise sphere, they are anticipated to replace HDDs.

Nonetheless, the rapid advancement of the flash memory market depends on the long-term establishment of a robust ecosystemIt is widely recognized that while flash memory hardware has progressed at an accelerated rate in recent years, the distinctive characteristics of Chinese market users, encompassing aspects like operating systems and processors, underscore the importance of software innovation and ecosystem development for the widespread adoption of new flash technologies and products.

Ni elaborates that the complexities inherent in the Chinese application environment for data centers necessitate a collaborative effort from all ecosystem participants to promote flash technology

To illustrate, the successful acceleration and application of QLC SSDs necessitate a concerted push from cloud computing customers, enterprise clients, OEMs/ODMs, and operating system vendors alike.

For example, many data center users contend with various service architectures involving Intel, AMD, ARM, IPU/DPU, and GPU, all of which face challenges related to mixed workload optimization, storage capacity and performance utilization, and write amplificationTo address these complexities, Solidigm has implemented its CSAL software to optimize workloads in data centers by effectively transitioning write workloads to sequential writes, substantially enhancing storage efficiency.

“CSAL has been open-sourced, allowing everyone within the ecosystem to adopt it, thereby further accelerating the proliferation and application of flash memory,” Ni shared.

Furthermore, the rise of AI presents a distinct opportunity for the update of data center infrastructure