TSMC considers making 3D DRAM chips at Nvidia’s request
Nvidia has asked TSMC to produce 3D DRAM memory chips, a technology that could ease chronic shortages in AI hardware. If TSMC agrees, it would become the first new major supplier in decades, challenging Samsung, SK Hynix, and Micron in a market where prices have doubled due to AI demand.
Editor, Lazyfounder

30 SEC SUMMARY
- Nvidia has asked TSMC to explore producing 3D DRAM chips, a technology that could ease chronic shortages in AI memory.
- TSMC’s entry would introduce a fourth major supplier to compete with Samsung, SK Hynix, and Micron in a market where prices have doubled due to AI demand.
- TSMC vice president Hsu Kuo-chin is leading the assessment, but no timeline or factory plans have been finalized yet.
- Micron expects memory shortages to last until 2028, prompting companies like Apple and Elon Musk to raise prices or plan new chip plants.
- Nvidia has also urged Samsung, SK Hynix, and Micron to build custom high-bandwidth memory using TSMC’s 3nm process.
TABLE OF CONTENTS
Why Nvidia wants TSMC in the memory game
Nvidia wants TSMC to start making 3D DRAM memory chips, and TSMC is now weighing the request. These chips stack memory cells vertically, a design that squeezes more capacity into the same space—a critical advantage for AI workloads that devour data.
How TSMC could break into the memory market
Hsu Kuo-chin, the TSMC vice president who oversees advanced packaging, is leading the internal assessment. The company has not committed to a timeline or a factory, but it is already studying two options: constructing a new plant in Taiwan’s Central Taiwan Science Park or converting an existing one. Either path would pit TSMC against Samsung, SK Hynix, and Micron, the three companies that have dominated memory production for decades.
Demand is swallowing supply whole. Micron, the largest U.S. memory maker, expects shortages to stretch into 2028, pushing prices higher. Apple raised prices on some Macs and iPads in June, citing soaring memory costs, while Elon Musk is planning a $55 billion chip plant to reduce his dependence on Micron.
Nvidia’s backup plan for memory supplies
Nvidia’s strategy extends beyond 3D DRAM. In October, it asked Samsung, SK Hynix, and Micron to manufacture the base layers of its next-generation high-bandwidth memory using TSMC’s 3nm process. The request acts as a safety net: if Samsung or SK Hynix stumble, TSMC’s advanced node could keep Nvidia’s GPUs flowing.
The move also strengthens Nvidia’s hand in price negotiations. "If you can credibly threaten to bring a third supplier into the game, you don’t have to accept whatever Samsung or SK Hynix quote," a semiconductor analyst told The Next Web.
The stakes for TSMC and the AI industry
TSMC’s internal review is still underway, but the broader industry is already moving. AI data centers are consuming DRAM at record rates, and regulators in multiple countries are scrutinizing chip consolidation. Meanwhile, OpenAI’s abrupt policy shifts on custom silicon show how quickly supply-chain assumptions can collapse.
For TSMC, the choice is clear. Staying out of memory keeps costs low but risks losing lucrative orders if shortages drag on. Entering the market costs billions upfront but could secure demand from Nvidia and others for years.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
TSMC’s potential move into 3D DRAM would give Nvidia and other AI chipmakers a critical second source for memory, breaking the trio of Samsung, SK Hynix, and Micron that has controlled the market for years.
Key takeaways
- TSMC’s entry could force Samsung, SK Hynix, and Micron to accelerate innovation or lower prices to retain customers.
- Nvidia’s request for custom HBM built on TSMC’s 3nm process could reduce its dependence on Samsung and SK Hynix for high-bandwidth memory.
- The shift reflects a wider industry panic: companies are racing to diversify suppliers before AI demand outstrips memory production.
- Even if TSMC never ships a single 3D DRAM chip, its review alone signals how seriously the memory bottleneck threatens AI growth.
Related on Lazyfounder
Sources
- The Next Web · 2026-11-06
TSMC is weighing a move into AI memory chips at Nvidia’s request
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.


