In the era of rapid evolution of the global storage industry and the blowout development of AI applications, AI DRAM has become a new strategic highland. As a key memory manufacturer in Taiwan, Nanya Technology is seeking breakthroughs in this red ocean market with unprecedented technology investment and product strategy. Its layout is not only an attempt to make a leap at the enterprise level, but also has the potential to leverage structural changes in the global DRAM ecosystem.
Ⅰ AI-driven new storage demand, and the market landscape needs to be reshaped urgently
The rapid development of AI is profoundly changing the demand structure of the storage industry. Traditional general-purpose memory can no longer meet the stringent requirements of AI in terms of high bandwidth, low latency, large capacity, and low power consumption. For example, AI training model parameters are often tens of billions, and real-time data analysis and inference also pose extremely high challenges to storage throughput.
According to Omdia, the global AI DRAM market size will be about $8.9 billion in 2023 and is expected to exceed $18 billion by 2026, with a compound annual growth rate of nearly 27%. In the face of this market with huge potential, major storage manufacturers have increased their layout. Samsung, SK hynix and Micron are still dominant, but the "second echelon" manufacturers, represented by Nanya Technology, are trying to cut in and compete for the right to speak through differentiated technologies.
Ⅱ Nanya Technology's four core layouts: directly hit the pain points of AI DRAM
Nanya Technology is focusing on the four major technology directions to build its differentiated competitive advantage in the field of AI DRAM.
1. High-density advanced DRAM: Break through the bottleneck of capacity and energy efficiency
As AI models grow in size, the demand for DRAM capacity soars. Nanya Technology continues to promote the development of 1Anm and 1Bnm nodes, and accelerate the transition to DDR5 and LPDDR5 generations. Its latest-generation 16Gb DDR5 DRAM chip has a 33% higher capacity density per unit area and a 25% lower power consumption compared to the previous generation, leading the industry in energy efficiency.
This high-density, low-power DRAM product greatly meets the sensitive requirements of AI for computing power and energy consumption ratio, especially in applications such as data centers, edge servers, and in-vehicle intelligent platforms, and has significant competitive advantages.
2. 3D TSV+ multi-chip packaging: Break through bandwidth bottlenecks
Faced with the limitations of traditional 2D packaging in terms of bandwidth and energy consumption, Nanya Technology has joined forces with Taiwanese packaging and testing manufacturers (such as Formosa Advanced Technology and Core-Patching Technology) to vigorously promote 3D TSV (through-silicon via) process and multi-chip integrated packaging (MCP) technology.
Through the TSV vertical interconnection structure, Nanya Technology has achieved a performance leap of doubling the chip stack height and increasing the IO bandwidth by 3\~4 times. At present, the transmission rate of its sample module has exceeded 1024 Gbps, which can be used for high-end inference servers and large-scale training platforms.
With the MCP package, the system integration is further improved, the product reliability is enhanced, and the memory solution for complex AI systems is flexible and scalable.
Figure: Can Nanya Technology's AI DRAM break the inherent pattern of the storage market?
3. HBM design capability: impact the high-end track
High-bandwidth storage (HBM) is currently monopolized by Samsung, SK hynix, and Micron, and the threshold is extremely high. However, Nanya Technology did not avoid this challenge, but chose to start from the design side and accumulate technical potential energy.
It is reported that it has set up an internal HBM design project team of more than 50 people, and has cooperated with world-renowned design service institutions. In terms of HBM controller architecture, signal integrity optimization, thermal management technology, etc., Nanya Technology has achieved preliminary results and is preparing for mass production in 2026\~2027 in the future.
If a technological breakthrough can be achieved in the field of HBM, Nanya Technology is expected to become another major force in the global AI storage market.
4. Logic Base Die: Building the cornerstone of intelligent memory
In the AI DRAM module, Logic Base Die is responsible for the core control and logic operation tasks, and is the key interface between DRAM and external systems. Nanya Technology actively cooperates with wafer foundries (such as PSMC and VIS to develop logic base chips with low power consumption and high integration).
Its latest samples were successfully taped out on the 5nm process, and the test results showed that the control delay was reduced by 18% and the energy consumption was reduced by 22%, which has the potential to enter the mainstream AI DRAM product line.
Ⅲ Customized strategy and market rhythm: 2026 is a critical time window
Nanya Technology's AI DRAM customization project will be verified in 2026 and start contributing revenue. Unlike general-purpose DRAM, AI applications require a high degree of customization in terms of bandwidth, capacity, power consumption, and packaging. Nanya Technology plans to collaborate with AI chip customers and system manufacturers to provide a "semi-standard + customized" product line.
According to the Global Semiconductor Industry Association (SIA), by 2026, the market share of custom AI DRAM will account for more than 25% of the total AI DRAM, far exceeding the current 10%. Nanya Technology's customized services can not only improve customer stickiness, but also improve ASP (average selling price) and gross profit margin, bringing more sustainable growth momentum to the company.
Ⅳ Seize the opportunity to transfer orders and strengthen short-term performance
During the adjustment period of the DRAM industry cycle, Samsung, Hynix, and Micron have reduced the production capacity of old DDR3/DDR4 products, causing some customers to look for alternative suppliers. Nanya Technology took advantage of the momentum to undertake the transfer of orders, and the shipment volume in Q1 2025 increased by 17% year-on-year, and the revenue reached NT$1.56 billion, an increase of 12% year-on-year, significantly easing the inventory pressure.
It is expected that the inventory depletion will be completed in the third quarter of 2025 and achieve a profit inflection point in the fourth quarter. This process will provide valuable cash flow and technical resources to invest in its AI DRAM project.
Ⅴ There are three major challenges that cannot be ignored
Despite the promising future, Nanya Technology still has multiple challenges to address:
1. High technical threshold: HBM and 3D TSV involve multiple difficult links such as design, manufacturing, packaging and testing, and still rely on external cooperation resources at this stage, making it difficult to be fully independent in the short term.
2. Market oligopoly pattern: Samsung, Hynix, etc., relying on a strong patent pool and production capacity layout, still have an overwhelming advantage in the short term;
3. New storage competition: Non-volatile emerging technologies, including MRAM, PCM, RRAM, etc., are on the rise, which may pose a substitution threat to DRAM in specific AI scenarios.
Ⅵ Summary: Breakthrough and future prospect in the challenge
On the whole, Nanya Technology's layout in AI DRAM is not to "copy" the giant route, but to try to break through a technological breakthrough in the gap between giants through four core technology breakthroughs, flexible product strategies and keen market timing.
If its HBM and custom DRAM projects are successfully implemented in 2026 and achieve breakthroughs in product performance, power consumption and cost, Nanya Technology will not only open a new revenue growth curve, but also may rewrite the pattern of the global AI storage industry.
In the next few years, it will be a key window to determine whether it will be among the mainstream in the global AI storage ecosystem. This unique battle of Nanya technology deserves the continuous attention of the industry.
