Huang Renxun personally went to TSMC to ‘demand more chips’ in response to storage shortages, price increases, and other issues

NVIDIA CEO Jensen Huang this week made a high-profile visit to Taiwan Semiconductor Manufacturing Company’s (TSMC) cutting-edge 3nm production line, underscoring the deep manufacturing partnership that has helped power the rise of AI hardware. Huang not only toured the facility, but also made a rare appearance at a TSMC employee sports day, publicly thanking the company and saying, “Without TSMC there would be no NVIDIA,” comments that highlight the critical interdependence between the chip designer and the world’s largest contract manufacturer.

The visit comes as NVIDIA ramps production of its next-generation Blackwell family of GPUs, which Huang described as experiencing “very strong” demand. The Blackwell architecture is anticipated to serve as the backbone for large-scale AI training and inference workloads, driving heavy orders for advanced-node chips that only a handful of foundries are equipped to produce. TSMC chairman and CEO C.C. Wei was quoted with a lighthearted nod to the pressure the orders place on Taiwan’s factories: “Huang asked for more chips,” a comment that captures both the boisterous tone of the meeting and the supply-side realities.

Industry watchers say the trip signals NVIDIA’s intent to secure priority access to scarce 3nm capacity. TSMC’s 3nm process is widely viewed as a key enabler for performance and power-efficiency gains essential to next-generation AI accelerators. The fabs’ limited initial capacity, meanwhile, has been a focal point for customers competing to lock in production slots. Huang’s public show of gratitude — and his visibility among TSMC staff — serves both as a strategic gesture and a reassurance to partners that NVIDIA is committed to long-term collaboration with its primary foundry.

When pressed about broader memory and storage market dynamics, including recent shortages and price fluctuations that have affected the industry, Huang was circumspect. He told reporters that decisions on pricing and supply were for memory manufacturers to make themselves, saying companies should “decide how to run their own businesses.” His response avoided a direct commitment from NVIDIA on whether it would intervene or adjust purchasing strategies to alleviate upward pressure on memory prices, which can materially affect overall system costs for AI deployments.

Market analysts note that while NVIDIA’s immediate focus is on securing advanced logic-node capacity for GPUs, memory availability remains a critical element of system-level performance and cost. Tightness in DRAM and NAND supply chains could compound pressure on customers seeking to deploy high-memory systems for AI inference and training.

The visit also serves as a reminder of the geopolitical and logistical complexities facing the global semiconductor industry. TSMC’s Taiwan-based manufacturing footprint is central to many global technology roadmaps, but capacity constraints and rising demand from AI leaders like NVIDIA are intensifying competition for scarce resources. For now, Huang’s visit and public praise reinforce the close alliance between NVIDIA and TSMC, even as other parts of the silicon supply chain — particularly memory — continue to raise questions about pricing and availability.