Nanya Technology is preparing a historic leap in its manufacturing capacity. The Taiwanese manufacturer plans to allocate over 200 billion New Taiwan Dollars, approximately $6.2 billion, to investments during 2027—nearly four times its budget for this year. The plan follows an exceptional quarter, with revenue soaring by 684% and a gross margin of 79.5%, boosted by the sharp increase in conventional DRAM memory.
Key Facts of Nanya’s Record-Breaking Year in 20 Seconds
- Planned Investment for 2027: over 200 billion New Taiwan Dollars, around $6.2 billion.
- Still Preliminary Plan: the budget has not yet received board approval.
- Q2 Revenues: 82.55 billion New Taiwan Dollars, up 684% year-over-year.
- Net Profit: 50.19 billion New Taiwan Dollars, a year-over-year increase of 1,324%.
- Gross Margin: reached 79.5%, compared to a negative margin of 20.6% in the same quarter of 2025.
- New Factory: set to have an initial capacity of 30,000 wafers per month in 2028, with potential expansion to 45,000 wafers later.
- Shift Driven Mainly by Prices: Nanya is benefiting from shortages in DDR4, DDR5, and low-power memories, not from proportional volume growth.
- Not Yet Competing in Advanced HBM: its business remains focused on conventional DRAM, although it is developing a custom AI solution for edge computing.
The figures for Q2 are still unaudited but illustrate the rapid change in the memory market within a year. Nanya moved from reporting industrial losses and negative margins to achieving profitability levels more typical of scarce products rather than of traditional high-volume memory manufacturing.
Between April and June, the company generated 82.55 billion Taiwanese Dollars in revenue—more than its total sales in some recent fiscal years. Net profit reached 50.19 billion, nearly eight times what it earned throughout 2025, according to results announced during the earnings presentation.
This growth does not mean Nanya has multiplied its shipments of memory sevenfold. The main explanation lies in prices. In the first quarter of 2026, the average selling price of its products had already increased by over 70% compared to the previous quarter, while the delivered bit volume decreased by a single-digit percentage.
A Shortage of Conventional DRAM Benefits Nanya
The race for AI has led Samsung, SK hynix, and Micron to dedicate increasing portions of their resources to higher-value memories, especially HBM and server DRAM. This shift has reduced the availability of conventional products for computers, mobile devices, networking equipment, automobiles, and industrial electronics.
Nanya’s position is unique. The company holds about 2% of the global DRAM market and does not manufacture HBM generations used in the most advanced GPUs. Around 70% of its shipments in early 2026 consisted of DDR4 and LPDDR4, with DDR5 accounting for roughly 10% of revenue.
In another phase of the cycle, exposure to mature products would have been a disadvantage. Now, it has become a source of profitability. Major manufacturers have been reducing DDR4 capacity to focus on DDR5, HBM, and data center products, but millions of servers, computers, integrated devices, and industrial systems still require the previous generation.
Supply has declined faster than demand, and buyers unable to redesign their equipment immediately are accepting significantly higher prices. Nanya has operational production capacity and can capture part of this market without currently entering the technical complexity and investments needed to directly compete with SK hynix’s, Samsung, or Micron’s HBM.
TrendForce forecasts that contractual prices for conventional DRAM will rebound by 13% to 18% in Q3 2026. The rise would be smaller than earlier in the year but would extend a period of extraordinary margins for manufacturers.
Nanya’s CEO, Pei-Ing Lee, predicts that tighter supply constraints will persist through the first half of 2027, possibly extending the imbalance into 2028. Expanding capacity requires building cleanrooms, installing lithography equipment, and completing validation processes—none of which can be accelerated at the same pace as price increases.
The company expects to increase its bit shipments by low-double digits over the course of 2026. This indicates that revenue growth will continue to depend more on what each customer pays than on the physical amount of memory produced.
This situation also carries risks. The DRAM market is known for its volatile cycles. When new capacity ramps up or demand softens, prices can fall rapidly, turning previously lucrative investments into factories with lower profitability.
A $15 Billion Factory for DDR5 and Low-Power Memories
Most of the 2027 budget will go toward accelerating Nanya’s new factory in Taishan District, New Taipei City. The project will require about 480 billion New Taiwan Dollars, roughly $15 billion, once fully operational.
Its first phase is expected to produce 30,000 wafers per month in 2028. Later, the facility could reach 45,000 wafers per month, still below the capacity of Samsung or SK hynix’s largest fabs but sufficient to transform Nanya’s current scale.
Initially, the factory will use the company’s 1B node, its second-generation DRAM process at 10 nanometers class. It will produce DDR5, DDR4, and low-power variants for computers, servers, mobile devices, and other applications.
The over-2027 budget of more than 200 billion Taiwan Dollars contrasts sharply with investments in recent years. Nanya spent 13.2 billion in 2023, 16.1 billion in 2024, and 13.4 billion in 2025. For 2026, it budgeted up to 52 billion.
This planned single-year investment exceeds the sum of the previous four years. Pei-Ing Lee clarified that this figure is preliminary and must be approved by the board, so it could still change depending on equipment prices, construction pace, and market conditions.
Nanya strengthened its balance sheet before embarking on this endeavor. In March, it closed a private capital increase of approximately 78.72 billion Taiwan Dollars (around $2.5 billion) subscribed by SanDisk, Kioxia, Solidigm, and Cisco. In exchange, these four groups collectively received about 10.19% of the Taiwanese manufacturer.
Their investment is not solely financial; SanDisk, Kioxia, and Solidigm produce SSDs which rely on DRAM for caching and internal table management. SanDisk and Kioxia also established long-term supply agreements with Nanya.
Solidigm’s involvement is especially notable since it is owned by SK hynix, the second-largest global DRAM manufacturer. A subsidiary choosing a smaller competitor highlights how buyers are seeking to secure capacity and mitigate the risk of component shortages.
HMM Margins Without Manufacturing HBM
The gross margin of 79.5% approaches levels achieved by producers that sell high volumes of HBM used in AI accelerators. In Nanya’s case, almost all of this profit derives from conventional DRAM, making this quarter a particularly clear snapshot of pricing pressures.
The company has ruled out direct entry into generations like HBM2, HBM3, HBM3E, or HBM4—markets demanding advanced processes, chip stacking, complex packaging, and close collaboration with GPU designers.
However, Nanya is working on a custom memory solution oriented toward AI applications in peripheral devices. This project is developed jointly with Etron Technology, Piecemakers Technology, and Formosa Advanced Technologies, with an aim to introduce its first prototype by late 2026.
It would not directly compete with memory used by Nvidia or AMD accelerators in large data centers. Its target applications include systems where higher bandwidth, lower power consumption, and cost efficiency are critical—such as robots, smart cameras, vehicles, industrial equipment, or compact servers.
This strategy allows Nanya to avoid direct competition with companies investing billions in HBM, but it also exposes the company to the market that has indirectly supported its recent recovery. AI absorbs advanced production capacity from competitors, leaving less available for the types of memory Nanya can produce.
The challenge is to use current profits to modernize its technology before the cycle shifts. Achieving and maintaining margins of 79.5% might be unsustainable if prolonged, potentially inviting new capacity additions.
The Taishan factory will not produce its first 30,000 wafers per month until 2028. By then, prices, AI demand, and the balance between HBM and traditional DRAM could all have evolved.
Nanya is seizing a rare opportunity: earning extraordinary profits in a segment the industry’s biggest players had begun to consider secondary. The 2027 investment will decide whether this temporary benefit allows expansion in size and technology or if it arrives too late, after shortages have eased.
Frequently Asked Questions
How much will Nanya invest during 2027?
The company plans capital expenditures exceeding 200 billion Taiwan Dollars, roughly $6.2 billion. This budget remains preliminary and awaits final approval from the board.
Why has Nanya’s revenue surged so much?
Primarily due to the increase in average DRAM selling prices. Revenues rose 684%, but the volume of memory shipped did not grow proportionally.
Does Nanya manufacture HBM memory for AI?
Currently, it does not produce HBM2, HBM3, HBM3E, or HBM4. Its focus is on DDR4, DDR5, and low-power memories, though it is developing a custom AI solution for edge computing.
When will the new factory start production?
The first phase is expected to achieve a capacity of 30,000 wafers per month by 2028 and expand later to 45,000 wafers.
via: Nanya
