NVIDIA's Capital Strategy: Investing Across the Entire AI Industry Chain

Since the AI boom began, NVIDIA has been regarded by most as the “pickaxe seller”—if you want to train a model, you need to use its cards. But recently, from acquiring Hugging Face to investing $3.5 billion in MediaTek, NVIDIA has been putting real money on the table, embedding itself into every link of the AI industry chain. From the fourth quarter of 2025 to early September 2026, NVIDIA’s publicly disclosed investments and acquisitions have covered at least four layers: entry points and key technologies, cutting-edge model companies, GPU cloud infrastructure, and chips and supply chain. The deals take the form of mergers and acquisitions, strategic equity investments, convertible bonds, investment rights, and technology licensing.The most eye-catching are the acquisitions. In 2019, NVIDIA bought Mellanox for $7 billion and basically took a break from major acquisitions over the following years. On September 3, 2026, it officially announced the acquisition of open-source AI platform Hugging Face for $12.93 billion, seizing control of the largest distribution entry in the open-source model ecosystem.In addition, its transactions with Groq and Poolside mark a new approach invented by NVIDIA in this round of capital activity: “licensing + talent acquisition + equity participation,” bypassing cumbersome regulatory approvals and reacting swiftly to industry trends.NVIDIA’s layout in cutting-edge models is even broader: its name appears on the funding lists of OpenAI, Anthropic, SSI, and xAI. CFO Colette Kress confirmed during the earnings call at the end of August that NVIDIA has invested nearly $50 billion in frontier AI laboratories.At the computing power infrastructure layer, NVIDIA’s investment logic is somewhat different. For mature GPU cloud platforms that already have significant orders, NVIDIA chooses to buy equity to secure the relationship. For projects still in the expansion phase, it prefers to lock in future rights first, then add investment once the project matures.Within the chip and hardware supply chain, NVIDIA’s investments extend into unexpected areas: investing in optical fiber giant Corning to secure fiber capacity for the next-generation Rubin GPU cluster’s high-speed interconnections; purchasing shares in Synopsys and Nokia; and investing in MediaTek through convertible bonds.As for emerging sectors like embodied intelligence and generative content, these are managed by NVIDIA’s venture capital division, NVentures. The focus is not on single investment amounts but on bringing representative startups into NVIDIA's ecosystem early with small equity stakes.Besides spending money, NVIDIA is also making money back via long-term computing power contracts. Investment, procurement, and long-term agreements work together to extend one-off equity relationships into multi-year demand for computing power.Most recently, NVIDIA is driving something even more disruptive: transforming computing power from one-off equipment procurement into infrastructure assets that can be financed, operated, and configured long-term. It is mobilizing over $500 billion in third-party Wall Street capital towards AI data center projects; its partnership initiative with South Korea’s SK Group also exceeds $500 billion, covering 2GW AI factories and long-term HBM memory contracts.NVIDIA has never intended to just sell batches of GPUs; what it truly aims to construct is a capital narrative around AI computing power as fundamental infrastructure.