Nvidia May Make Its Own Memory to End Samsung Reliance 2026

Nvidia is reportedly considering producing its own advanced memory to reduce dependence on Samsung and other memory makers, according to a leak reported by The Information. The move comes as the AI boom has triggered a severe memory shortage that has sent prices soaring and is expected to last until at least 2028.

Nvidia, one of the main drivers of this crisis, has grown tired of relying on just a few manufacturers for the memory used in its graphics cards. The company has therefore explored making its own advanced memory, at least according to the leak.

What the Leak Says About Nvidia's Memory Plans

The report, published by The Information, revealed that TSMC had begun drawing up plans to manufacture 3D RAM at Nvidia's request. TSMC has its own roadmap, but Nvidia is the Taiwanese company's 'Customer A' and, in the current situation, clearly has the leverage to make such requests.

The outlet cited two sources with direct knowledge of the matter, and Reuters picked up the information, although it could not independently confirm it. According to the initial report, TSMC could buy a factory and convert it into a plant for producing memory, similar to a move already made by companies like Micron, which specializes in memory.

The report also indicated that the project would be led by Hsu Kuo-chin, a vice president who oversees advanced packaging at TSMC and was previously an executive at Micron in charge of its Taiwan operations.

Why Nvidia Is Considering Making Its Own Memory

The AI boom has created unprecedented demand for high-bandwidth memory (HBM) and other advanced memory types used in GPUs. Supply has struggled to keep up, leading to skyrocketing prices and allocation challenges for Nvidia and its partners.

By producing its own memory, Nvidia could secure supply, reduce costs, and gain more control over its product roadmap. However, building memory fabrication capabilities is a massive undertaking that requires specialized expertise and billions of dollars in investment.

TSMC's involvement makes sense given its close relationship with Nvidia and its advanced packaging capabilities. The company already handles CoWoS packaging for Nvidia's GPUs, which is a critical step in the manufacturing process.

Key Takeaways

• Nvidia is reportedly exploring in-house production of advanced 3D RAM with TSMC's help.
• The move aims to reduce reliance on memory suppliers like Samsung amid a severe AI-driven memory shortage.
• TSMC would potentially convert an existing factory for memory production, led by a former Micron executive.
• The shortage is expected to persist until at least 2028, driving up costs across the tech industry.
• Nvidia's position as TSMC's largest customer gives it significant influence over such projects.

How the Memory Shortage Affects the Tech Industry

The memory shortage is not just an Nvidia problem; it affects the entire tech ecosystem. From smartphones to data centers, advanced memory is a critical component, and the AI boom has exacerbated existing supply constraints.

Memory manufacturers have been slow to expand capacity due to the high costs and long lead times of building new fabs. This has led to a situation where demand far outstrips supply, giving memory makers significant pricing power.

Nvidia's potential move into memory production could disrupt this dynamic. If successful, it would give Nvidia more leverage over its supply chain and potentially lower costs for its products.

Challenges and Implications

Producing advanced memory is not easy. It requires deep expertise in semiconductor manufacturing, and TSMC would be entering a new area if it takes on this project. The company's core strength is logic chips, not memory.

However, TSMC has a history of successfully entering new markets, and its advanced packaging capabilities are closely related to memory production. The involvement of Hsu Kuo-chin, with his experience at Micron, adds credibility to the plan.

If Nvidia does start making its own memory, it could reshape the competitive landscape. Memory makers like Samsung, SK Hynix, and Micron could face a new competitor in their largest customer.

What to Watch Next

For now, the plan is just a leak, and neither Nvidia nor TSMC has confirmed it. The companies are unlikely to comment on rumors about unannounced products. The background to this story is set out in NVIDIA is considering the return of the Samsung Foundry-produced GeForce RTX 3060. A related thread runs through Nvidia GPUs: Hyperscalers Hoard Billions in Unused Chips. Earlier coverage in Strait of Hormuz Blockade Threatens Samsung and TSMC Chip Supply shows how the argument developed.

Investors and industry watchers should keep an eye on TSMC's capital expenditure plans and any announcements about new fabs or acquisitions. Nvidia's supply chain moves will also be closely watched as the AI boom continues. A closer look at the same subject appears in PC companies are considering postponing new products due to memory availability. Anyone following the wider picture can start with Nvidia CEO Jensen Huang Reportedly Targeted by Chinese Spy. The same theme surfaces in Sony and TSMC Joint Venture to Make Advanced Mobile Sensors, which takes a different angle.

The memory shortage is expected to persist for several years, so any move to increase supply will be significant. Whether Nvidia goes it alone or partners with TSMC, the company is clearly looking for ways to secure its future. Context on how this played out before sits in Pokemon Cards Match Nvidia as Top-Performing Asset. A companion piece, Apple Briefly Tops Nvidia as Most Valuable Company, approaches the same subject from another angle.