New Delhi: Shares of Taiwan Semiconductor Manufacturing Company (TSMC) surged more than 6 per cent in US trading on Friday, November 6, after reports emerged that Nvidia had asked the world’s largest contract chipmaker to explore manufacturing advanced memory chips. The development could mark a significant expansion of TSMC’s role in the artificial intelligence (AI) hardware industry, taking it beyond its established business of manufacturing processors.

According to a report , TSMC has started drawing up plans to manufacture a next-generation memory technology known as three-dimensional dynamic random-access memory, or 3D DRAM. The proposal reportedly came at the request of Nvidia, TSMC’s largest customer.

The report cited two people familiar with the discussions. Reuters also reported the development, while noting that the companies had not confirmed the proposed manufacturing plans.

The possibility of TSMC entering the memory market attracted investor attention because the company could become a new competitor to established suppliers such as Samsung Electronics, SK Hynix and Micron Technology. These companies currently dominate the global DRAM industry.

Why TSMC shares jumped more than 6 per cent

TSMC’s US-listed shares rose more than 6 per cent in early trading on November 6 following the report. The rise reflected investor optimism about the possibility of the company expanding into another important segment of the AI hardware supply chain.

TSMC is already a crucial manufacturing partner for Nvidia. It produces advanced processors used in the company’s AI accelerators, which power data centres and artificial intelligence applications.

However, powerful processors need equally advanced memory to store and retrieve data quickly. The rapid expansion of AI data centres has increased demand for both computing chips and high-performance memory, creating supply constraints across the semiconductor industry.

If TSMC eventually begins producing advanced DRAM, it could supply another important component required by AI hardware manufacturers. The move could also give Nvidia an additional potential source of memory as it seeks to secure supplies for its growing range of AI products.

The share-price movement reflects market expectations about this potential opportunity. It does not mean that TSMC has already entered the memory business or secured a manufacturing agreement with Nvidia.

What is 3D DRAM and how does it work?

DRAM, or dynamic random-access memory, is a type of semiconductor memory that temporarily stores information while a computer or processor is operating.

Conventional DRAM stores memory cells across a relatively flat structure. As manufacturers attempt to increase capacity and improve performance, shrinking individual components becomes increasingly challenging.

3D DRAM aims to address some of these limitations by stacking memory cells vertically in multiple layers. This architecture could allow manufacturers to increase memory density within a given footprint and potentially improve the efficiency of data storage and access.

The technology is attracting attention as AI workloads demand increasingly large quantities of memory. Advanced AI systems need to move substantial amounts of information between processors and memory, making memory capacity, speed and power consumption important factors in overall performance.

However, 3D DRAM is not the same as high-bandwidth memory (HBM). HBM combines multiple memory dies into vertically stacked packages to provide high data-transfer capacity. 3D DRAM, as discussed in the reported TSMC proposal, involves arranging the memory cells themselves in a three-dimensional architecture.

TSMC already participates in the advanced packaging ecosystem used for AI processors and HBM-based products. Manufacturing 3D DRAM would represent a further step into the memory chip business itself.

Nvidia’s search for more memory supply

Nvidia’s reported request comes at a time when the AI industry is facing intense demand for advanced memory.

The company’s graphics processing units (GPUs) are widely used in AI data centres. These processors require high-performance memory to handle the enormous datasets and calculations involved in training and running artificial intelligence models.

The rapid construction of AI infrastructure has increased demand for memory products, including HBM and conventional DRAM. Major memory manufacturers are under pressure to expand production capacity while meeting the requirements of several large technology companies.

Nvidia’s interest in TSMC could help it explore another potential source of supply in the longer term. It could also strengthen the relationship between the two companies, which already work closely on advanced processor manufacturing.

However, any additional memory production from TSMC would take time to develop. The reported plans do not indicate that new chips will become available immediately or help resolve current shortages.

TSMC could build new factories or acquire an existing facility

The reported proposal is still under consideration, and TSMC has not announced a final manufacturing strategy.

The company is examining two possible routes. One option would be to construct new semiconductor factories dedicated to memory production. Another would involve acquiring an existing manufacturing facility and converting it for DRAM production.

Building a new fabrication plant would require substantial investment, specialised equipment, technical expertise and several years of development. Acquiring an existing facility could potentially reduce the time required, although conversion and production preparation would still involve considerable work.

The reported evaluation is being led by Hsu Kuo-chin, a TSMC vice-president responsible for advanced packaging operations.

TSMC has experience in memory manufacturing from earlier decades. However, the company eventually concentrated on its contract foundry business, producing chips designed by other semiconductor companies.

A return to memory manufacturing would therefore represent a notable strategic change.

Samsung, SK Hynix and Micron could face a new competitor

The global DRAM market is currently dominated by three major manufacturers: Samsung Electronics, SK Hynix and Micron Technology.

These companies have established manufacturing capabilities, technical expertise and relationships with major electronics and technology businesses. They are also investing in next-generation memory technologies to address growing demand from AI applications.

TSMC’s potential entry could increase competition in a market where supply has become strategically important.

However, moving from research and planning to commercial-scale memory manufacturing is a difficult process. Producing advanced DRAM requires high manufacturing yields, reliable technology and significant capital expenditure.

The memory industry is also cyclical. Periods of shortages and rising prices can encourage manufacturers to increase production, but additional capacity may eventually create oversupply if demand slows.

That risk could influence TSMC’s decision on whether to invest in a new business that operates differently from its established foundry model.

TSMC has not confirmed a production timeline

Despite the market reaction, the reported expansion remains uncertain. TSMC told The Information that it does not comment on market rumours or speculation and reiterated that its core business is dedicated integrated-circuit foundry manufacturing.

Nvidia did not respond to the publication’s request for comment, according to the report.

There is currently no confirmed production start date, investment amount or factory plan for the proposed 3D DRAM business. The company would need to resolve these questions before the project could become a commercial manufacturing operation.

For now, Nvidia’s reported request has placed TSMC at the centre of another major development in the AI semiconductor industry. If the company eventually moves ahead, it could expand its position from manufacturing AI processors to supplying another critical component of the systems powering artificial intelligence.

The immediate share-price rally reflects investor interest in that possibility. Whether the proposal becomes a major new business for TSMC will depend on its investment decision, manufacturing capabilities and the future economics of the global memory market.