Nanya plans $6.2B spend in 2027, a 4x jump on AI chip demand

Key Takeaways

• Nanya will spend T$200 billion ($6.2B) in 2027, roughly four times its 2024 budget
• Q2 revenue jumped 684% year-over-year with gross margin swinging from -20.6% to 79.5%
• A new plant targets 30,000 wafers/month by 2028, expandable to 45,000 wafers/month

Taiwan's Nanya Technology plans to spend more than T$200 billion ($6.2 billion) in 2027, roughly quadrupling its current budget, as AI-fueled demand for memory chips rewrites the economics of the DRAM industry. President Pei-Ing Lee disclosed the preliminary figure during a Friday press briefing, calling it a ramp-up for a new fabrication plant that will anchor the company's growth for the next decade.

The numbers behind the announcement tell their own story. Nanya reported unaudited second-quarter revenue of T$82.55 billion, a 684% surge from the same period last year. Net income climbed 1,324% to T$50.19 billion. Gross margin flipped from negative 20.6% a year ago to 79.5%. Those swings reflect the violent rebound in memory pricing after a brutal down-cycle, now supercharged by datacenter and AI accelerator demand.

What is Nanya building with T$480 billion?

The centerpiece is a new fabrication plant that will absorb roughly T$480 billion in total investment once it reaches full production capacity. Phase one is scheduled to hit 30,000 wafers per month by 2028, with an eventual ceiling of 45,000 wafers per month. This year alone, Nanya expects to spend over T$50 billion, mostly on early construction and equipment procurement.

Lee noted that the T$200 billion figure for 2027 remains a preliminary target and still requires board approval. Even so, announcing a number of that magnitude signals confidence. Nanya's customers include Nvidia, Qualcomm, and Google, all of whom are burning through memory at unprecedented rates to train and serve large AI models.

Why is AI reshaping memory economics?

High-bandwidth memory (HBM), stacked DRAM used in GPUs like Nvidia's H100, has become the bottleneck in AI infrastructure. But the ripple effects extend to commodity DRAM too. Larger clusters of servers need more standard memory for host CPUs, storage caching, and networking gear. Lee said structural changes driven by AI are supporting a stronger long-term outlook for the memory industry and expects the current supply shortage to persist for several more quarters.

Nanya is Taiwan's third-largest DRAM producer, behind Micron's Taiwan operations and local rival Powerchip. It has historically focused on commodity memory for PCs and consumer electronics, a segment more exposed to cyclical swings. The pivot toward AI-adjacent capacity is a strategic hedge and a growth bet rolled into one.

How does Nanya compare to Samsung and SK Hynix?

Global leaders Samsung Electronics and SK Hynix have both announced massive capex increases of their own. SK Hynix dominates the HBM market, supplying the stacked memory for most of Nvidia's AI accelerators. Samsung is racing to catch up after quality issues slowed its HBM3 ramp. Nanya's $6.2 billion outlay is smaller in absolute terms, but the 4x year-over-year jump signals comparable urgency.

Commenting on South Korea's push to expand semiconductor production, Lee called such efforts positive for the broader industry and a reflection of confidence in market demand. The subtext: Nanya is not intimidated by Korean scale. It sees room for every capable supplier in a market where AI demand is outstripping available wafers.

What are the risks?

Memory is notoriously cyclical. The same industry that delivered 79.5% gross margins this quarter ran at negative margins just twelve months earlier. A slowdown in AI infrastructure spending, an inventory glut, or a recession could send prices back down. Nanya's new plant, if it comes online during a downturn, would add supply into a weak market.

Geopolitical risk is another variable. Taiwan sits at the center of semiconductor manufacturing, and any disruption to its fabs ripples across the global supply chain. Nanya's concentrated footprint amplifies that exposure.

Frequently Asked Questions

Why is Nanya increasing spending by 4x in 2027?

AI-driven demand for memory chips has pushed pricing and margins sharply higher. Nanya is investing T$200 billion ($6.2B) to build a new fab and expand capacity to meet orders from major customers including Nvidia and Google.

When will Nanya's new plant reach production?

Phase one targets 30,000 wafers per month by 2028. Full capacity of 45,000 wafers per month will follow, with total investment reaching T$480 billion.

How did Nanya perform in Q2 2024?

Revenue rose 684% year-over-year to T$82.55 billion. Net income surged 1,324% to T$50.19 billion. Gross margin flipped from negative 20.6% to positive 79.5%.

Who are Nanya's main competitors in DRAM?

Samsung Electronics, SK Hynix, and Micron dominate global DRAM. Nanya is Taiwan's third-largest producer, specializing in commodity memory for PCs and consumer electronics.

What risks does Nanya face with this expansion?

Memory pricing is cyclical. A demand slowdown, inventory buildup, or recession could erode margins. Geopolitical risk from Taiwan's location adds another layer of uncertainty.

Source: Tech-Economic Times / ET