Nvidia

Nvidia's stated positioning is the complete datacenter: GPUs, software and networks together. Cobus Greyling argued in October 2025 that the real shift was into software, with open-source Nemotron-Nano-12B-v2-VL-FP8 models framed as the backbone of scalable agentic systems, the DGX Spark ARM64 workstation as a desktop entry point for researchers, and fine-tuning methodology captured through the tooling. Greyling put the lock-in plainly: "No easy AMD/Intel swaps, but that's the point right?"

Tariq Krim read the hardware position as an infrastructure duel: Nvidia GPUs as a de facto standard against Google's TPU and integrated ecosystem, set inside a capital realignment he refused to call a bubble. A Bloomberg analysis from October 2025 described capital circulating between OpenAI and Nvidia across a $1 trillion market, with OpenAI not projected cash-flow positive before the end of the decade.

Dependency is the European angle. Mistral AI's gigawatt build plan lists Nvidia GPUs among its unresolved dependencies. Arthur Mensch described Campus IA at Saint-Arnoult (35 Md€, 100 hectares, 1.4 to 1.6 GW) as an MGX and Nvidia project in which Mistral AI is a very minor shareholder. BCG files L'Oréal's alliance with Nvidia under partnership rather than acquisition, L'Oréal's own advantage resting on its biological data about skin and hair.

Sasha Levin, kernel developer and Nvidia engineer, announced the Maintainers Summit consensus to deprecate the Linux kernel's Assisted-by tag in October 2026. A BFM Business panel reported Nvidia paying bonuses in tokens.