NVIDIA is reportedly facing gigantic demand for its Hopper H200 AI chips from Chinese customers, and the firm is now prompting TSMC to take 'desperate' supply chain measures.
NVIDIA Now Sees Demand For H200 AI Chips That Is 2x Higher Than Its Current Inventory
We have reported in the past on how newer solutions introduced in China by NVIDIA/AMD are garnering massive attention within Chinese hyperscalers, mainly because the region is in desperate need of compute power to fuel advancements in frontier AI models. According to a report by Reuters, NVIDIA has received orders for up to 2 million H200 chips for next year, while the company's current inventory stands at just 700,000 units. This suggests that NVIDIA and its supply chain partners will likely need to restart Hopper production, placing significant constraints on companies like TSMC.
The moves raise concerns over whether there could be further tightening in global AI chip supplies as Nvidia now has to strike the right balance between meeting robust Chinese demand and addressing constrained supplies elsewhere.
- Reuters
NVIDIA's reliance on TSMC for foundry needs is biting them right now, given that the Taiwan chip giant is already facing bottlenecks in addressing demand for Blackwell and related products from hyperscalers worldwide. More importantly, the primary constraint for TSMC right now might not be semiconductor production, since the H200 does utilize TSMC's 4nm node, which is being produced in both Taiwan and the US, but the main bottleneck comes from CoWoS packaging, since the technology has been dominantly adopted by Hopper, Blackwell, and Blackwell Ultra products.
Reuters estimates that the ASP of an H200 AI chip in China is around $27,000, which means that for a two-million shipment figure, you could estimate a revenue of $54 billion from China, far exceeding what NVIDIA had written off earlier when the export restrictions were implemented. The demand from Chinese customers is something that NVIDIA cannot ignore at present, but a more pressing question is whether the AI supply chain can actually meet the efforts needed to satisfy global demand. TSMC already faces an explosion in CapEX and labor shortages, yet partners are demanding more from the Taiwan giant.
H200 AI chips are known to be six times more powerful compared to the H20 in training workloads, which is one of the reasons why China's AI industry is rushing to place orders. It would be interesting to see how the NVIDIA-China story evolves moving forward, as the current demand will induce massive pressure on the supply chain.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
