October 13, 2025
About the Company
Nanya is a small Taiwanese DRAM manufacturer. They produce DRAM on trailing-edge process nodes (licensed from Micron) but are developing their own more advanced, technology. Nanya has around 1% share of the DRAM market, a distant fourth, behind Samsung, Hynix, and Micron. The top three companies control more than 90% of the DRAM market.
I don’t own any shares of Nanya and would not buy any, because they lack the scale and the technology to be competitive in DRAM. I keep an eye on their quarterly releases as a data point on the DRAM market.
Market Commentary and Q2 2025 Performance
Eleven quarters have passed since Nanya stopped including the time trend of their gross margin with their earnings. That is almost three full years. After increasing by almost 50% last quarter, Nanya’s revenue rose by an even more stunning 78.4% sequentially in the third quarters. This rise came from a combination of higher bit shipments (up mid-twenties %) and better average selling price (increased forties %). The market for Nanya’s legacy DRAM bits is benefitting from a combination of lower investment in non-leading-edge capacity, the end-of-life notifications on DDR4 and LPDDR4 products from the Big Three, and bits being drawn away from conventional DRAM as they are shifted to HBM production. The forties % increase (I’m estimating it to be a 43%) in pricing is the largest sequential ASP jump I’ve ever seen. This is only possible because the market for legacy DRAM is a minor part of the overall market, and thus can experience more severe swings. It should be noted that the three quarters before this one all saw Nanya’s ASPs decline. Going back to the beginning of the non-HBM DRAM upturn (the HBM upturn began two quarters earlier), Nanya’s average selling price per bit is up a little more than 50%. That is a CAGR of 24.1%. Over that same two-year stretch, the company’s bit shipments are up by around 75%, close to a 33% CAGR. Industry estimates are that the DRAM market will grow at a mid-teens percent CAGR over the medium to long term. Pulling the lens back further, using the last calendar quarter of 2021 as a starting point – nearly four years ago – Nanya’s bit shipments have increase by low-50s percent. Said another way, the company today ships 50% more bits than it did at the peak of the last upturn, fifteen quarters ago. For comparison, over the same fifteen quarter period rose by half, Hynix’s DRAM shipments also rose by 50% and Micron’s rose by 90%. So over that nearly-full cycle (3.75 years), the DRAM bit growth CAGR for both Nanya and Hynix has been around 12%. Micron’s bit CAGR has been almost 19%. A key difference between Hynix and Nanya over that time is Hynix has been converting DRAM wafers over to HBM production at a 3:1 trade ratio. Nanya has no HBM technology. Their bit growth has been almost all from adding wafer capacity and tech node migration. I suspect some of their most recent growth has been from drawing down inventory, taking advantage of the severe shortages in the market. Nanya’s gross margin swung positive again, clocking in at 18.5% Operating margin was 6% and net margin was 8.3%, up from negative 39% last quarter. The business made a complete 180 turn this quarter.
Bit Shipments and Capital Expenditures
The company has spent NT$11.2B of capital expenditures in the first three quarters of calendar 2025. Last quarter, the company said their board had approved total spending of NT$19.6B. They have down revised that total to NT$16.0B. This is surprising given how hot the market is right now. However, the investment reduction is not affecting bit growth for the remainder of this calendar year. For the fourth quarter in a row, they have increased their 2025 year-over-year bit growth target. After Q4 of last year, the company said they would grow bits by 20% in 2025 over 2024. After Q1 the said this would be 30%. Following Q2 they said bit growth would be up 40%. Now, they have raised their forecast for sequential bit growth in 2025 from 2024 to 50%. The reduction in capital expenditures may be taking bits out of 2026. It may be because of more efficient use of capex dollars. We’ll learn more at the end of the year when they give their first guidance for 2026. The 50% bit growth in 2025 over 2024 is massive.
DRAM Market Outlook for Q4 of 2025
Strong AI demand and constrained non-AI bit supply are fueling the current DRAM market strength. Both AI and non-AI servers are expected to drive DRAM demand growth in 2026. The company also thinks AI adoption is expanding to non-cloud locations such as at the edge. Major suppliers have shifted capacity from DDR4 and LPDDR4 to DDR5 and LPDDR5. Inventory at manufacturers remains healthy.
· Server Market: AI and HPC continue to drive HBM and DDR5 demand.
· PC & Mobile Market: New products use more memory, which is always the case, yet memory companies continue to point it out.
· Consumer Market: Demand is growing for DDR4 and LPDDR4 for consumer applications.
The company is trying to develop their own HBM product.
Analyst Call
Prepared Remarks
· Pei-ing Lee, the CEO, said last quarter that the company’s ASPs had declined in April and May of this year, then started to rise in June. This has carried through from June until today. He didn’t say this explicitly, but pricing being up “forties%” tells you this is true.
· The company sees “opportunities” for positive outlook in Q4-25 and the quarters beyond. Even by Nanya’s conservative commentary standard, this is an understatement, at least for the fourth quarter.
Analyst Q&A
· The PC sector will move to DDR5 faster than other sectors, in response to the higher prices in DDR4 seen today. In the consumer segment, DDR4 and LPDDR4 will continue to play a “major role.” This is because many of these applications need 8Gb parts and the DDR5 minimum is generally 16Gb.
· Their CapEx in 2026 will “stay low” because their new fab building won’t be ready until 2027. Output improvements in 2026 will need to come from manufacturing efficiency and yield improvements.
· The DDR4 shortage will continue for “some quarters.” PC demand will reduce over time but consumer demand will stay strong. The recent price increases they have seen and are seeing are coming from all memory types; DDR3, DDR4, DDR5, and derivatives of these architectures.
· They are currently preparing their second generation 10 nm product for high bandwidth memory. They intend to engage in the AI market. Claiming customer confidentiality, management declined to be specific about which AI product they are developing. From what Dr. Lee did say, it sounds like a high-bandwidth product targeted at the mobile or PC market.
· One analyst asked if 635K wafers is their current output, to which Dr. Lee responded that is reasonable. I think this means 635K per year, which is about 53K wafer outs per month. That is in the right range for Nanya. Dr. Lee also said they are space constrained right now. As they advance nodes, they lose wafer output because more advanced nodes take more floor space.
· Because of the shortage of memory, there is more customer interest in, and discussions about, long-term contracts. These discussions are ongoing. The DRAM business does not have a history of customers signing binding long-term take-or-pay contracts.
· About 60% of their output is consumer segment, 15-20% is LP/mobile, with the rest being PC and cloud. I didn’t realize they are this heavily weighted to the consumer segment.
· In 2026, their bit shipments will be within plus-or-minus 10% from where they are right now. They are already running at a high level of shipments.
· Their DDR4 and DDR5 inventories are healthy. DDR4 is in shortage.
· The equipment installation into their new fab is going to happen in 2027.
· Their new fab facility will start with the third and fourth generations of their 10nm process node.
Summary
Starting in June of this year, the legacy DRAM market moved into undersupply, which has now become acute. The company’s revenue rose sequentially by nearly 80%, an unheard of increase quarter over quarter. It is only possible to do this by taking a lot of bits out of inventory. Behind that inventory, the company is also ramping bit output from their fabs. The are now saying that bit output in 2025 will be 50% higher than it was in 2024. They have been raising their growth target since the beginning of the calendar year. Based on comments from the CEO during the call, their fabs have now hit the walls. Further bit output will have to come from increasing manufacturing efficiency and yield. Said another way, their current cleanroom space is completely full of equipment. Further evidence of this is in Dr. Lee’s forecast that 2026 bit output will be within +/- 10% of their output in 2025. As they have been ramping throughout 2025, I’m not sure if he means 2026 output will be flat from the rate they exit 2025. The other option, the one I think is the case, is that they are drawing down a lot of inventory to ship the number of bits they are selling today. The company is getting a stay of execution from the strength that has emerged in the DRAM market over the last two months. 60% of their bit output goes into the consumer segment. That is the lowest-margin segment in memory, but also the one that holds on to the legacy architectures that Nanya specializes in. They are making hay while the sun shines, making as many bits and clearing out as much inventory as they can in the current strong market.
– S. Hughes (short MU)