Nanya Technology’s NT$200bn capex bet tests whether AI has changed the DRAM cycle

Nanya Technology Corporation (TWSE: 2408) is considering more than NT$200 billion of capital expenditure in 2027, approximately four times its approved 2026 budget, to accelerate development of a new advanced-memory fabrication plant in Taiwan. The first phase is expected to reach 30,000 wafer starts per month during 2028, with the longer-term plan expanding capacity to 45,000 wafers per month under a total investment programme of about US$16 billion. Nanya Technology reported second-quarter revenue of NT$82.55 billion, net income of NT$50.19 billion and a gross margin of 79.5% as constrained supply drove a sharp rise in DRAM pricing. The shares last traded at NT$396 before Taiwan’s July 10 typhoon-related market closure, down approximately 2.7% from the July 2 close but still near the upper portion of a 52-week range spanning NT$40.20 to NT$505. The central investment question is whether artificial intelligence has created a structurally stronger memory market or merely given Nanya Technology enough cash to make an unusually large commitment near the peak of another semiconductor cycle.

Why is Nanya Technology planning NT$200 billion of 2027 spending at the top of a memory upcycle?

Nanya Technology’s preliminary spending plan is remarkable when compared with its recent investment history. The company spent NT$13.2 billion in 2023, NT$16.1 billion in 2024 and NT$13.4 billion in 2025. Its approved 2026 capital budget is up to NT$52 billion, meaning the proposed 2027 expenditure would exceed the combined capital spending of the previous four years.

The company is accelerating because available DRAM production is becoming increasingly valuable. AI servers require large quantities of high-bandwidth memory, DDR5 and other advanced memory products, encouraging Samsung Electronics, SK hynix and Micron Technology to redirect equipment and wafer capacity towards higher-margin applications. That shift reduces the capacity available for personal computers, smartphones, automobiles, consumer electronics and conventional servers.

Nanya Technology can benefit even without dominating the highest-performance segment. When larger manufacturers prioritise high-bandwidth memory, shortages can develop in DDR4, low-power DDR4 and standard DDR5 products that remain essential across the wider electronics economy. Nanya Technology can sell into those gaps while gradually increasing its exposure to newer products.

The investment also reflects the unusually long construction and equipment timetable for semiconductor factories. Waiting until shortages become permanent would be too late because clean rooms, manufacturing tools, utility systems and process qualifications can take several years to complete. The 2027 spending surge is intended to make meaningful capacity available from 2028 rather than address only the current quarter’s demand.

The difficulty is that memory producers have repeatedly expanded during periods of exceptional profitability, only for new capacity to reach the market after demand weakened. Nanya Technology is betting that AI, edge computing and higher memory content per device have altered that historical pattern enough to justify a much larger fixed-cost base.

Are Nanya Technology’s 79.5% gross margins sustainable or mainly a pricing windfall?

Nanya Technology’s second-quarter results show the extraordinary effect of constrained supply on a commodity semiconductor business. Revenue rose 68.2% from the first quarter and 684.2% from the prior-year period. Gross profit almost doubled sequentially to NT$65.62 billion, while operating income increased to NT$60.83 billion.

The key driver was pricing rather than production growth. Average selling prices rose by more than 60% from the first quarter, while bit shipments remained broadly flat. This means the increase in revenue and profit came primarily from customers paying far more for each unit of memory rather than Nanya Technology producing substantially more chips.

A 79.5% gross margin is exceptional for a conventional DRAM manufacturer. It reflects scarcity economics rather than a normal long-term return profile. Nanya Technology’s gross margin was negative 20.6% only one year earlier, illustrating how quickly earnings can move when supply and pricing change.

The current profitability provides a valuable window for investment. Nanya Technology generated NT$55.01 billion of operating cash during the second quarter and ended June with NT$216.76 billion of cash and equivalents. Net cash was approximately NT$198.4 billion after accounting for long-term debt, giving the company substantial capacity to fund construction without immediately relying on excessive borrowing.

Investors should nevertheless separate the strength of the balance sheet from the durability of the margin. If average selling prices normalise while depreciation, staffing and utility costs increase, returns from the new fab could fall rapidly. The investment case depends less on preserving a 79.5% margin and more on ensuring the new plant remains competitive when margins return to less extraordinary levels.

The strongest sign of sustainable improvement would be rising bit shipments alongside strong pricing. The current quarter delivered record economics on flat volume, which is excellent for cash generation but less informative about how the company will perform after additional supply enters the market.

How can Nanya Technology benefit from AI demand without leading the mainstream HBM market?

Nanya Technology is not positioned as a direct high-bandwidth memory leader comparable with SK hynix, Samsung Electronics or Micron Technology. Those companies are spending heavily on advanced stacking, packaging and manufacturing technologies required to supply HBM alongside AI accelerators.

Nanya Technology instead benefits from the capacity displacement created by HBM. High-bandwidth memory consumes significant manufacturing resources and requires more complex production than conventional DRAM. As major suppliers allocate wafers, capital and engineering talent towards HBM, customers in traditional memory markets face tighter availability.

This creates a second-order AI opportunity. Nanya Technology can capture demand for DDR4, DDR5 and low-power memory used in servers, networking equipment, solid-state drives, personal computers, automobiles and industrial devices. AI infrastructure itself also requires conventional DRAM across central processors, management controllers, networking systems and storage devices.

AI infrastructure and server applications now contribute more than 20% of Nanya Technology’s revenue. That exposure is meaningful even though the company is not supplying the most visible memory installed beside flagship graphics processors.

The company is also developing customised AI and wafer-on-wafer products, including memory intended for edge and specialised computing applications. This may provide a route into physical AI, robotics and embedded systems where customers require lower power consumption, predictable performance or application-specific designs rather than the largest possible HBM stack.

This positioning can reduce direct competition with the three dominant HBM suppliers. It may also limit Nanya Technology’s access to the highest-margin portion of the AI memory market if conventional DRAM supply becomes more balanced.

Nanya Technology’s strategy is therefore based on breadth and scarcity rather than technological leadership in one flagship category. That can be commercially effective, but it requires management to increase DDR5 and customised AI exposure before customers eventually obtain sufficient conventional memory from larger competitors.

Why does the new Taiwan fab matter for DDR5, edge AI and global supply diversification?

The first phase of Nanya Technology’s new fabrication plant is planned to reach 30,000 wafer starts per month in 2028 and later expand towards 45,000. The facility will support the company’s transition into newer process generations and a broader portfolio spanning DDR5, low-power DDR5, DDR4 and other memory categories.

The plant is important because Nanya Technology’s long-term competitiveness cannot depend indefinitely on shortages of mature products. Customers will continue migrating towards DDR5 and newer low-power memory as processors, servers and consumer devices adopt updated architectures.

Newer manufacturing processes can produce more memory bits from each wafer, reduce power consumption and improve product economics. They also require substantial equipment spending and process-development capability, particularly as the industry introduces extreme ultraviolet lithography into advanced DRAM production.

Nanya Technology is developing its 1C, 1D and 1E process generations and expects future EUV capability to support the roadmap. Successful execution would narrow part of the technology gap with larger rivals, although Samsung Electronics, SK hynix and Micron Technology will continue investing at much greater absolute scale.

The project also provides supply-chain diversification. Nanya Technology manufactures in Taiwan, giving global customers an additional source outside the dominant South Korean and United States memory producers. Diversification has become strategically important as AI demand, export controls and geopolitical tensions expose the risks of relying on a small number of suppliers.

However, Taiwan concentration creates its own risks. Semiconductor production requires stable electricity, large amounts of water, specialised chemicals and uninterrupted logistics. Natural disasters, infrastructure constraints or cross-strait tensions could affect both the existing operations and the new fab.

The facility’s value therefore depends on more than installed wafer capacity. Nanya Technology must achieve competitive yields, secure equipment, qualify products with customers and operate the fab efficiently enough to remain profitable when memory prices eventually weaken.

What do the SanDisk, Kioxia, Solidigm and Cisco investments reveal about customer strategy?

Nanya Technology completed an NT$78.72 billion private placement during April, giving four strategic investors approximately 10.19% of the enlarged share capital. The investors were SanDisk Technologies, Kioxia Holdings, SK hynix-owned Solidigm and Cisco Systems.

The transaction strengthens Nanya Technology’s balance sheet before the most capital-intensive phase of the new fab. Ending the second quarter with substantial net cash reduces refinancing risk and gives management greater flexibility to place equipment orders while market conditions remain favourable.

The strategic identities of the investors matter more than the cash alone. SanDisk, Kioxia and Solidigm manufacture storage products that require DRAM for caching, control and system performance. Cisco Systems supplies networking infrastructure where memory content is increasing alongside AI, cloud and data-centre workloads.

SanDisk and Kioxia also entered long-term DRAM supply arrangements with Nanya Technology. These agreements provide customers with greater visibility over future availability while helping Nanya Technology align capacity investment with expected demand.

The structure resembles a partial customer-financed expansion model. Customers gain equity exposure and supply security, while Nanya Technology gains funding and stronger commercial relationships before committing billions of dollars to new production.

This reduces, but does not eliminate, demand risk. Strategic investors may purchase significant volumes during a shortage and reduce requirements later if inventories rise or technology changes. Long-term agreements can contain pricing, volume and flexibility provisions that affect the value ultimately retained by Nanya Technology.

Existing shareholders also experienced dilution. The transaction can create value if the capital funds high-return capacity and the partnerships support durable sales. It would be less attractive if the company issued equity near the peak of the cycle only to invest in capacity that later produces weaker returns.

Can Nanya Technology narrow the manufacturing gap with Samsung, SK hynix and Micron?

Nanya Technology is competing against three companies with much larger research budgets, customer relationships and production networks. Samsung Electronics, SK hynix and Micron Technology have established positions across conventional DRAM, advanced server memory and HBM.

Nanya Technology does not need to match their full scale to create value. A smaller supplier can remain strategically important when customers want alternative sources, specialised products and supply assurance. Even a modest share of a large and capacity-constrained market can support strong earnings.

Independence is also commercially useful. Storage, networking and device companies may prefer to source DRAM from a supplier that does not compete directly with them in NAND flash, solid-state drives or complete computing platforms. Nanya Technology can position itself as a neutral memory partner.

Its challenge is manufacturing cost. Larger competitors can spread research, process-development and equipment costs across more wafers. They can also shift capacity between product categories and negotiate better terms with equipment and materials suppliers.

China’s ChangXin Memory Technologies adds another competitive dimension. Chinese memory capacity is growing and could place pressure on mature DRAM pricing over time, although export controls and geopolitical concerns may restrict its role in some global supply chains.

Nanya Technology must therefore compete through a combination of technology migration, customer partnerships and disciplined capacity deployment. Simply producing more DDR4 during a shortage will not establish a durable position.

The company’s best route is to secure customers before capacity is installed, move a greater proportion of output into DDR5 and customised AI products, and maintain enough flexibility to adjust the production mix when market conditions change.

What does Nanya Technology’s July market performance reveal about investor expectations?

Nanya Technology last traded at NT$396 on July 9 because the Taiwan Stock Exchange was closed on July 10 as a typhoon affected the market. The closure means investors had not yet produced a normal trading-session reaction to the earnings and capital-spending announcement by the July 11 publishing date.

The NT$396 closing price was approximately 2.7% below the July 2 close of NT$407. The short decline should be viewed against a much larger rerating, with the stock still trading almost ten times above its 52-week low of NT$40.20.

The shares were also about 21.6% below their 52-week high of NT$505. That gap indicates that investors had already become more cautious after an extraordinary rally, even before the company proposed another large increase in capital spending.

The valuation debate is unusual because current earnings make the stock appear inexpensive on near-term profit measures. Those profits are being generated during an exceptional pricing environment, so investors cannot assume the latest quarterly earnings represent a stable annual run rate.

The market is likely to examine whether Nanya Technology’s 2027 budget receives board approval and how management phases equipment commitments. A staged programme would allow the company to adjust spending if demand weakens, while an aggressive fixed schedule could increase late-cycle risk.

The lack of an immediate post-announcement session is important for publishers and investors. The July 11 feature should not claim that the market welcomed or rejected the NT$200 billion proposal. The most recent share price reflects expectations formed before the full second-quarter presentation and press briefing.

Which execution risks could turn the NT$200 billion expansion into a costly late-cycle bet?

The first risk is board approval. Management has described the 2027 capital budget as preliminary, so the final amount, timing and allocation may change before formal authorisation.

The second risk is memory pricing. Nanya Technology’s earnings improved because average selling prices rose dramatically while shipments remained flat. A fall in pricing would reduce cash generation before the largest expenditure commitments are completed.

The third risk is manufacturing execution. Advanced DRAM processes require complex equipment, precise yield management and lengthy customer qualification. Delays or weak yields could leave the new fab carrying depreciation without producing sufficient saleable output.

The fourth risk is technology mix. Nanya Technology may add capacity for DDR4 and standard DDR5 while the highest-value market shifts faster towards HBM, low-power memory or integrated AI architectures.

The fifth risk is equipment availability. EUV systems, deposition equipment, etching tools and advanced metrology products are supplied by a limited number of companies. Delivery delays can affect construction schedules and increase costs.

The sixth risk is customer concentration. Strategic supply agreements improve visibility, but a large portion of new demand may become linked to a relatively small group of storage and networking companies.

The seventh risk is geopolitical exposure. The new fab deepens Nanya Technology’s manufacturing concentration in Taiwan at a time when global customers are seeking greater geographic resilience.

The eighth risk is environmental infrastructure. Advanced fabs require reliable power, water and waste-management capacity. Expansion can face community, regulatory or utility constraints even when the semiconductor market is strong.

The ninth risk is capital discipline. Record profits can make almost every expansion appear affordable. The true test arrives when prices decline and management must continue funding depreciation, maintenance and research without the benefit of scarcity-level margins.

What must Nanya Technology deliver before its expansion creates durable value per share?

Nanya Technology must first translate pricing-led revenue growth into higher production volumes. The 2026 bit-shipment target calls for growth in the high teens, which would show that earnings improvement is not entirely dependent on price inflation.

The company must also increase the proportion of revenue coming from DDR5 and customised AI-related products. A broader advanced-product mix would reduce dependence on shortages of mature DRAM and improve the long-term economics of the new fab.

Management should provide a staged 2027 spending framework separating construction, wafer-fabrication equipment, EUV investment and process-development costs. Investors need to understand which commitments can be delayed if the market changes.

Customer-backed demand must become visible through shipments rather than strategic announcements alone. The SanDisk and Kioxia supply agreements will be more valuable if they provide predictable volume and reasonable pricing across the cycle.

Nanya Technology should preserve a strong net-cash position during construction. The current balance sheet is a competitive advantage, but the proposed capital programme is large enough to consume liquidity quickly if operating cash flow normalises.

The company must also demonstrate competitive yields at its 1B process while keeping the 1C, 1D and 1E roadmaps on schedule. Capacity without cost-effective process technology would weaken returns and make the company more vulnerable to larger rivals.

The AI memory shortage has given Nanya Technology a rare opportunity to expand from a position of financial strength. Whether that opportunity becomes durable value will depend on spending before customers need the capacity without spending so aggressively that shareholders inherit the next industry glut.

Key takeaways on Nanya Technology’s NT$200 billion AI memory expansion strategy

• Nanya Technology is considering more than NT$200 billion of 2027 capital expenditure, roughly four times its approved 2026 budget.
• The new fab is expected to reach 30,000 wafer starts per month in 2028 and eventually expand towards 45,000 under a US$16 billion total plan.
• Second-quarter revenue rose to NT$82.55 billion, while gross margin reached 79.5% and net income increased to NT$50.19 billion.
• Flat sequential bit shipments show that the latest profit surge was driven primarily by a more than 60% increase in average selling prices.
• Nanya Technology benefits from AI because HBM production absorbs capacity that would otherwise supply DDR4, DDR5 and low-power memory markets.
• AI infrastructure and server products now account for more than 20% of revenue, even though Nanya Technology is not a mainstream HBM leader.
• Strategic investments from SanDisk, Kioxia, Solidigm and Cisco Systems provide capital, customer alignment and potential long-term demand visibility.
• Nanya Technology ended June with approximately NT$198.4 billion of net cash, giving it substantial but not unlimited capacity to fund the expansion.
• The stock’s extraordinary rise from its 52-week low means investors are already pricing a significant improvement in memory-market economics.
• Volume growth, advanced-product adoption, fab yields and disciplined spending will determine whether the expansion creates lasting value per share.