NVIDIA dependence hits 90% costs

Nvidia’s problem is not demand. Demand is absurdly strong. The problem is that more and more of Nvidia’s business now runs through the same Asian manufacturing bottlenecks(tomshardware.com)across follow-on coverage this week, pegs Asian suppliers at about 90% of Nvidia’s production costs, up from roughly 65% a year earlier. (tomshardware.com) ### Why did(connecttech.com)depends on the same familiar chain — TSMC for advanced fabrication, SK hynix and Samsung for high-end memory, and big Asian assemblers for server systems. But now robotics, automotive, and “physical AI” products are piling onto that same base. (finance.yahoo.com) ### What does “90% of production costs” really mean? Basically, the expensive parts of Nvidia’s hardware stack are increasingly tied to suppliers in Taiwan, South Korea, and broader Asian manufacturing hubs. This is less about where Nvidia books revenue and more about where the hard stuff gets built — advanced wafers, mem(tomshardware.com)in tightens, the effects spill everywhere. (msn.com) ### Where is the strain showing up first? One clear signal is Jetson. Nvidia and partners are accelerating end-of-life plans for several LPDDR4-based Jetson modules, including Jetson TX2 (finance.yahoo.com)customer notice from April 27 lays it out pretty plainly — this is a memory market problem, not just a routine product refresh. (connecttech.com) ### Why does that matter beyond embedded developers? Because Jetson sits inside real products with long lives — robots, industrial systems, edge boxes, vehicles. Those teams do not swap compute(msn.com)r module gets squeezed out early, the cost is not just buying a new board. It is redesign, software porting, retesting, and timeline risk. (digitalcitizen.life) ### What’s happening at the high end? At the other extreme, scarcity is turning Nvidia’s newest AI systems into regional luxury goods. Reuters reported on April 30 that B(connecttech.com) in the U.S. The jump came as export controls and a crackdown on smuggling dried up gray-market supply. (money.usnews.com) ### So is this a supply problem or a geopolitics problem? Both — and that is the catch. Some shortages are industrial, like older LPDDR4 becoming less available. Some are policy-driven, like B(digitalcitizen.life), and more pressure to optimize around what is actually available instead of what looks best on paper. (connecttech.com) ### Why is robotics suddenly part of this story? Because Nvidia wants robotics to be a major growth lane, and Jetson Thor is the clearest example. Nvidia announced general(money.usnews.com) one more fast-growing category competing for advanced process capacity and memory. (nvidianews.nvidia.com) ### Bottom line? Nvidia’s moat is still huge. Revenue is still exploding — fiscal 2026 hit $215.9 billion. But the company’s success is making its supply chain more crowded, more Asian-cen(connecttech.com)city somewhere in the stack and plan around it early. (investor.nvidia.com)