NVIDIA

NVIDIA Corporation is the world's most valuable semiconductor company, founded in 1993 by Jensen Huang, Chris Malakowski and Curtis Priem. Originally known for gaming GPUs, it has become the dominant provider of AI computing hardware, with its chips powering the vast majority of AI training and inference worldwide, and Huang one of the most influential figures in technology.

The product line runs from data center GPUs and the DGX and HGX systems built around them to NVLink and NVSwitch interconnects that make multi-GPU scaling possible. The deeper moat is CUDA, the software platform with millions of developers, because most AI software is written and optimized for it. NVIDIA has since extended beyond the accelerator itself — into CPUs purpose-built for agentic workloads, and into Arm-based laptop processors co-designed with MediaTek that pair a custom CPU with an NVIDIA GPU, putting the company into direct competition with Intel, AMD and Qualcomm in personal computing for the first time. Above the silicon sits a broad software ecosystem: NVIDIA AI Enterprise for production deployment, Omniverse for digital twins and simulation, Drive for autonomous vehicles and Clara for healthcare.

The supply chain is anchored in Taiwan, where TSMC fabricates NVIDIA's chips and Foxconn assembles them into server racks. Huang has committed the company to spending on the order of 150 billion dollars a year there and broken ground on a Taipei headquarters, framing Taiwan as the epicenter of the AI buildout. Memory is the tightest constraint on shipping accelerators rather than raw compute, which is why a multi-year supply agreement with SK Hynix, paired with a gigawatt-scale AI factory that SK Telecom will build in Korea, is strategically consequential.

NVIDIA has moved from selling the chips to underwriting the system around them, and this is the most distinctive thing about its current position. It is the AI sector's single largest strategic investor, anchored by its stake in OpenAI and extending through glassmakers, data-center operators and dozens of private rounds — a pattern analysts describe as circular, with money cycling between chip vendor, model customer and infrastructure provider. It has gone further into financing: letters of intent with six large capital allocators to mobilize third-party capital for AI infrastructure, backed by a residual-value backstop under which NVIDIA covers part of any shortfall when installed chips resell below expectations. The same logic reappears at the site level, where it has guaranteed land, power and buildings behind a multi-gigawatt campus in exchange for being the exclusive compute provider. A hardware vendor taking depreciation risk on its own installed base, in order to keep its customers financeable, is a change in kind rather than degree.

Two structural facts complete the picture. NVIDIA has effectively lost the Chinese AI accelerator market after years of export controls and Beijing's homegrown-stack push, with Huang conceding its share there is near zero and Huawei the principal beneficiary. And the company is moving up the stack out of silicon: it has agreed to acquire Hugging Face, the open-model hub used by millions of developers, for 12.9 billion dollars — a deal agreed rather than closed, with Huang committing that the Hub will stay open to the whole ecosystem and that NVIDIA compute will not be required to build or deploy through it.